All sensible, but at odds with what we believe to be true. It is generally believed among TPR people that, while each section is curved as you explain, the rules for 15s (only) are not as the statistical approach would dictate -- to get a 15 requires, in practice, a perfect score, regardless of test form or section. This belief is the reason I questioned an earlier post, which stated a different view, so intently.
I don't think many of us can prove this, as we usually don't get to see our raw scores, but in fact my own PS performance last August supports my view -- to be honest, I couldn't imagine having missed more than one question on the PS last time, and given the distribution of scores in PS I suspect that would have put me in the top 0.1% (my test form had a rather difficult PS section, relative to others), but I still recorded only a 14. (I guess it is possible to see one's raw score, isn't it? I didn't, but someone who did might be able to settle this.)
If my hypothesis is correct, 14 on the PS should be somewhat broader and 15 somewhat narrower, in terms of percentiles, than what is reported by AAMC.
This is all speculation, though based on a ton of anecdotal evidence as well as some insider knowledge (second- and third-hand, though). Anyone who knows for certain how the calculations work for scores of 15, please tell us. The only datum that will prove me wrong is a confirmed 15 with at least one question wrong in the corresponding section; unfortunately, there's no way to prove me right.
