[First] You raise many very interesting points and I've made comments about them below. In fact, there's too much of interest for me to fully address in just one reply (you'll note this long meandering reply has had to be extended into the next reply below to fit it).
It's a rare experience for me to come across someone who's done the kind of work you have. On the few occasions I have and I follow up with questions, the person usually clams up because of trade secrets or such—you know 'our chocolate is smoother because of our secret process' etc., etc., so I've never gotten to the bottom of the aspects of the subject that I've been most curious about.
I'm not a chemist but given the number of courses I've done where chemistry was one of compulsory subjects, I reckon I ought to be. ;-) Once, I added the years up that I'd studied the subject and it came to over a decade (that, of course, includes high school). If it hadn't been for considerable duplication of subject material due to the silly way courses were constructed then, by now, I ought to be expert in the subject. But alas that's not so.
Now back to your comments:
"If they are tasting it blind (with no preconception), their taste is accurate."
"We used the same nomenclature and tasting training as the wine industry. It's a qualitative assessment, but there is a lot of science behind it, and the training is actually quite difficult. You actually train on specific samples with a single taste, initially very unsubtle and clearly labelled. Then to pass certification you have to do it completely blind.
Next level up you progress to subtler tastes and repeat. The really advanced people are genuinely capable of discerning very specific chemical compounds with a very high degree of accuracy.
Now let me also include my facetious and somewhat unfair comment (unfair for reasons I develop below):
"… I'm sure people can be trained to develop wine-tasting skills and that they actually get better with experience—but ultimately one cannot make a silk purse out of a sow's ear no matter how good the training."
Further down, you'll become aware that the 'silk purse/sow's ear' comment becomes very relevant.
Before I proceed I must add that decades ago one of the 'sins' of my youth was to spend a little time working in a vineyard/winery—not as a winemaker or specialist but as a general dogsbody. Nevertheless, when knowledgeable people who regularly win awards for their wines surround one, one cannot but help pick up a few valuable tips.
OK, back to the comments. All these quotes have common threads, which are, (a) about how do you calibrate the taste and olfactory processes of different individuals so their 'output' is uniform and thus useful to others, and, (b) when exposed to the same stimulus how do different individuals differ in their experience of it. We're confronted with the ancient conundrum that's as old as philosophy itself—not to mention the subject of countless philosophical dissertations—some of which I once had to study. It boils down to this: I cannot experience a given physical phenomena and accurately remember it then jump into your head to see if you/your senses experience it the same identical way that I did when I was in my own body.
Of course, there are two issues here. The first is that one's physical sensors (of taste, smell etc.) have given (nominal) sensitivities and that these differ between individuals. Moreover, they change dynamically over time and with the type and strength of the input signals (tastes). These sensors also possess bandwidth and detection thresholds in the physical sense. One experiences (directly perceives) sensory data from these senses (as say in a blind tasting, or feeling something) but that's about as good as it gets because one can have errors directly associated with perception. For instance, at one end of the spectrum you've lost your glasses and you mistake a ballpoint for a pencil, at the other extreme we've the case of the man with visual agnosia that Oliver Sacks outlined in is book The Man Who Mistook His Wife for a Hat. Then there are interpretation errors (errors of thought/mental processes), which can be caused by any number of reasons from having a vivid imagination, to schizophrenia, to reacting badly to a normal everyday stimulus because of an ongoing memory of a similar bad experience that one has had in the past.
This can quickly become a real can of worms, as things can get so inextricably intertwined and complicated. Where do plain old errors of perception end and interpretation errors actually begin? Let's take an extreme case, say someone is on LSD and they see or feel ants crawling up their arms or spot flying pigs then we have to ask what is at fault? Has the physical sensory system been corrupted and the data it's issuing is garbage, or is it that the sensory data is OK and that mental interpretation turns out junk (like a RAM fault), or is it a combination of both? The modern work/interpretation of this stuff essentially began with the sense-data constructs/notions of B. Russell and G.E. Moore et al, ca 1900, now much of it rests in the hands of neurologists.
Likely, these matters all sound rather hypothetical; the trouble is that the combined 'product' or output of these processes has real consequences in that they distort our perceptions of reality. In practice, they regularly play a large role in our day-to-day experiences. Essentially, they collectively add unwanted distortion to the output (that's to say our composite perception of an event is often in error).
For example, as a young kid I once grossly over-indulged in chocolate and made myself physically sick, afterwards I never ate another skerrick of it for over 20 years (as it made me feel sick whenever I tried ) [In hindsight, it was likely a damn good outcome.] Thus, emotion, likes, dislikes and experiences all play a huge role in interpretation.
The dynamics of all of this are seemingly overwhelming—perhaps confusing is a better word. So for the moment, let's reduce this to the simplest scenario (as per your "If they are tasting it blind (with no preconception), their taste is accurate." case. If I eat a sweet chocolate then immediately taste wine, it will taste bitter (even so a sweet sauterne); if I've just eaten a complex curry then my taste threshold changes (usually by desensitising/increasing the perceptible detection threshold), moreover, it will do so in a non-linear way (that's my experience anyway).
For example, take the case where I've not eaten (i.e.: my palate is 'clean') and (a) I dilute a sugar-water solution to the threshold to where I can just detect the sweetness, and (b) I dilute a water-denatonium (Bitrex) mixture to the point where I can just detect its foul bitterness. Then eating just about anything will alter those thresholds (usually by desensitising them but not always)—and the extent of this desensitisation will have occurred in a non-linear way. That's to say the thresholds will not desensitise to the same extent for each type of taste. If I then add a taste bias, add sweetness, bitterness, etc. into what I eat next, then the detection threshold for each chemical will change yet again. Moreover, it will again do so in a non-linear manner (this further compounds the effect).
Similarly, my taste bandwidth is reduced (here, I use the term 'bandwidth' to mean the extent to which I can differentiate varying degrees of sweetness/bitterness—essentially the range of different chemicals that I can detect for given concentrations of each). [Excuse my limited nomenclature here, my profession is IT and electronics, so it's those terms that make sense to me.]
Essentially, all of our senses have inbuilt Automatic Gain Controls, Automatic Bandwidths and Automatic Thresholds. And all of our various senses dynamically adjust or 'recalibrate' their thresholds to different detection levels depending on (a) the amount or level of the input stimulus/stimuli and (b) the intrinsic characteristics of those stimuli (type of chemical, etc.).
(Here, I've used both singular and plural deliberately, a singular stimulus will solicit a specific response, multiple stimuli will interfere with one another and thus give another different and separate response (as with intermodulation distortion in electronics, the resulting products (new tastes) of the combined responses will then be different to their singular counterparts. Furthermore, it's unlikely that the taster will be able to separate out the new taste(s) into their previous constituents—that is unless he/she does so from experience and or memory.)
Moreover, the sensitivities of our various senses and their recovery times (i.e.: the default input state after a given period of rest after in which no further stabilisation occurs) vary both over the short term (minutes/hours/days) and long-term (years/decades). We know this as we've many well documented cases. For example, there are thousands of instances where soldiers have been wounded on battlefields and their otherwise extremely severe pain is suppressed to tolerable levels—often for days afterwards. My understanding is that battle is such a terrible, emotional and cathartic experience that their perceptions of pain recalibrate as a coping mechanism. Years or even decades later when the same soldiers experience even a much lesser pain for whatever reason in normal civilian circumstances then they do so to the same extent as everyone else. Essentially, they now experience a considerably great pain for a lesser stimulatory input than they did for their battlefield injuries (sometimes this perplexes them greatly, as they worryingly perceive that they've now lost the courage and stamina that they once had.)