If we take your suppositions, let’s assume 90% of the 17% who applied to medical school had GPA’s in excess of the Harvard median of 3.8, which is over 15% of the total class. So of the remaining top half of the class, 35% of total students were spread among top law schools, finance, consulting, tech (FB/Google/etc…), pharma/research, plum government/NPO positions and top graduate programs? Or looking at it another way, almost 31% of the top 50% applied to med schools? Doesn’t seem probable. I think it is more probable that top half of the Harvard class had a wider dispersion of post graduate outcomes.
But we don’t even need to go there. The UCB grads with 3.9+ GPA and 515+ MCATS only had an admit rate of 87% (91/105). By the stats you stated, with relatively lower grade inflation at Berkeley (3.9 more select at UCB vs a Harvard 3.9), the Berkeley outcome for the highest stats group was lower than Harvard’s rate of 93% for a much wider group.