Apple is wealthy and powerful, and gets the cream of the crop - i.e. the first batch of the new 5nm output. This isn't new, apple always releases their models a few months before equivalent snapdragons come out. This is a meaningful advantage, to be sure, but if you're trying to figure out how much apple's perf win is due to their great chips and not merely a generational gap, it's misleading. Comparing the s10 today, which is clearly on a inferior process to the iphone 12 pro today isn't useful
from this perspective. It's better to compare the upcoming 5nm 888 to apples A14.
And here, the numbers just don't back up the claim that apple is running rings around qualcomm in terms of performance. Not just that; the difference between apple+qualcomm chips hasn't changed all that dramatically over the years; if anything it's gotten smaller.
Yes, if you look at power at equivalent performance that looks better for apple, but then that's always the case because nobody runs client side CPUs at an optimal perf/watt power, but faster. It's also not really how people use their phones, is it? People don't turn off the high-power cores in iphones; they use em as designed, i.e. at approximately equivalent power, just faster.
The point here isn't that Apple's win doesn't exist; simply that it's not the mythological night-day difference it's sometimes painted to be. Furthermore, because apple consistently releases earlier on a new process node, at the moment of their release the perf numbers look better than they are on average over the products lifetime; in effect, you're seeing the combination of both Apple's good CPUs, and a 1 generation gap. Then there's the fact that Apple's CPU's are so fast due not solely to the skill of their silicon division, but also their willingness to spend for more transistors. We don't have numbers for the 888 yet, but it's very likely going to be a lower-cost design if history is any guide (and the anandtech article goes into some other non-obvious ways in which is saves costs too, like power planes, that I bet apple is willing to go the extra mile on).
All that matters, because you're trying to predict how other CPUs will perform, and thus need to tease out the impact of stuff like process node, and apples willingness to spend extra on larger caches, and x86 vs ARM - and just ascribing all that perf uplift to apples magic pixie dust won't help you make predictions regarding the rest of the CPU market in the future.