Nice though experiment, but doesn't hold up to scrutiny on many levels (not meant in a condescending way, just to spinn that thought experiment further)

First, we can't even agree on how/what to exactly measure with penis length. Even if we fixed that, the actual distribution itself doesn't have buckets (like size ranges) but would contain all data points individually. If we measure with unreasonably high precision and assume that penis length is a real (in a mathematical sense) number than there is a 0% chance that two people have the same penis length and there would be no person that has an exactly average penis length.

Stepping back from this thought experiment, there is data for penis length and the standard deviation is wider than most people would expect, it's far from 90% being average unless your average bucket is something like 5-20cm