Of course the incoming class isn't homogenous. But that doesn't change the fact that almost everyone in the class has always been above average up to that point.
In order to have any reasonable degree of confidence that you're going to end up above median at your school, I think you probably have to have both GPA and LSAT above median, and one of them above 75. Not that those dictate 1L success, but they're the only objective and statistically correlated factors you have that aren't tainted by optimism bias.
This part sounds like the old joke where a guy's looking for his lost car keys near a streetlamp, and a buddy asks him "I thought you lost your keys in the club?", to which the searcher responds "yes, but I'm looking here because the light's better". You can't say "I know the correlation between uGPA/LSAT and 1L grades is only 0.4, but it's the only objective, quantifiable dispassionate data we have, so we're going to treat it as dispositive, or nearly so."
To use another example, the median LSAT score for takers my age is 144. I beat that by about 25 points, not because I'm a genius or an LSAT god (I'm neither), but because other than the dispassionate, quantifiable criteria of age, I'm probably in a different situation than most of the group I'm being compared to. Just because the researchers chose to segregate the takers by age, that doesn't magically normalize all of the other non-quantifiable, "soft" factors.
Simply put, I question the value of historical aggregate data when it comes to predicting an individual's
future performance. It's a tautology that only half of the class will be above median, but I don't believe that makes it true that every individual in the class has an exact 50% chance of landing above median.