It is worse than you probably think because of the fertilizer, transportation of crops (and operating farm machinery), drying of crops, fermentation to ethanol, and (perhaps most brutally) distillation to high enough grade to mix with hydrophobic gasoline.
I was doing reviews of at least a half dozen studies of this that dated back to the mid '00s, back then the most optimistic studies were around 1.3 joules of ethanol for 1 joule for extraction (which excludes the sunlight). Meaning you're burning 1 joule of gasoline to farm and transport and manufacture 1.3 joules of ethanol. But the pessimistic estimates from then were also below 1, meaning that you literally use more petrochemicals to support growing the crop than the crop produces in finished fuel.
Whether it's breakeven, a little positive, or a little negative we've known for a long time that it's marginal at best, thus the interest in cellulosic ethanol. Whether it's 0.7 joules out or 1.3 joules out per joule in depends on assumptions like how far you need to transport things and how dry the crops start. There's no way to avoid that uncertainty.
That whether it's positive, unity, or negative is sensitive to these variables is a statement of just how marginal it is. It should just be positive, always, or what are we doing? Confusingly, some oil wells are barely breakeven because they use so much energy to build the well and process the fuel.
You can follow the citations inside this article, but, for instance:
> Resulting EROI is typically around 1.4-1.5 [for shale oil].
So every 1.4-1.5 joules of shale oil you burn required 1 joule of other energy to extract. Net it produces only 0.4-0.5 joules of free energy after extraction, and that's... well, arguably not actually enough to operate a civilization but let's not go there.
https://en.wikipedia.org/wiki/Energy_return_on_investment#Sh...
Corn ethanol has been a scam and known to be one for a very long time. Shale oil is marginally less of a scam, but is truly hard to justify compared to PV solar or wind which both easily exceed 4 and 34 depending on the details.
But no, variation in the estimated EROEI is entirely normal and expected because different sites have different details.