You're not accounting for a fundamental bias in that line of thinking.
The players who deviate from the status quo are already more critical-thinking oriented, and therefore are better players on average than those not thinking critically game-to-game. Therefore these players, even if making an item choice that would lower their winrate by deviating from the normal build, may actually win more due to being better players.
A Sylas building Ludens thinking to kill the enemy ADC faster because the Jhin was 2/0/0 in the first 5 minutes is already better in his critical thinking and more likely to win regardless of item choice, even if that Luden's purchase is incorrect over Everfrost, for example.
Did you account for this bias?
That's an interesting thought, with that said it's very hard to measure that or estimate it. There is definitely some non-0 bias there, but I wouldn't have the expertise to really estimate how much that matters. My gut tells me that over a large enough sample size (which we tend to have) it mostly washed out to less than 1% of a total winrate increase. But my guess there is likely as good as yours.