This applies to Michigan specifically. They are willing to make academic sacrifices for the sake of their athletic programs and (maybe) a well-rounded student body. Not all schools operate this way.
For example, suppose the academic ability of jocks is normally distributed about 105 with a std dev of 15, whereas non-jocks are normally distributed about 100 with the same std dev. If the cutoff for admission is 110, but athletic talent gives you a 20 point bonus, then you're your comparing mean aptitude in a sample of non-jocks, given that every observation is >= 110, to the mean aptitude in a sample of jocks, given that every observation is >= 90. Under those conditions, you'll find a 10 point difference in favor of your sample of non-jocks, even though the population mean for jocks is higher.
I think you alluded to this point in an earlier reply, but it's worth spelling out.