Ah, yes...the old "AAMC tables are internally inconsistent" problem, sorry.
In general I take the percentiles over the percents given, as I think there are some rounding issues contributing to the percent column (0.051%+0.051% turning into 0.1% and 0.1%, for example) and the percentile column is less prone to that than adding up multiple rounded numbers.
However, I always forget about the discrepancy between the percent and percentile column in these tables - generally if something is 99.9th percentile, you
cannot have 0.2% of the takers scoring at or above that level.
AAMC seems to do some variation of percentile where they include your score in the percentile...aka instead of reading it as "you scored better than 99.8% of people" it should be "99.8% of people scored as well or worse than you"/"you scored worse than 0.2% of people"
So, yeah...reading it that way I'd look at the numbers for a 39 (99.5 percentile) and say that 0.5% scored 40+, which turns into 475 people, which could fill 2 HMS's.
Still, do you think that all people with a 40+ apply to HMS, have a strong gpa, and also don't apply to any other school which they might rather attend? I'm actually curious here. I would
not think that, but then, I went to an undergrad full of exactly the kind of people who avoided Harvard as an undergrad (one of my friends even matriculated there, then got into my school and went 'see ya!' after getting her Harvard student email and everything).