They could just make the median 32.
Does having the median around 500 make it more numerically pleasing? Having a median at 32 and then a normal distribution around that ranging from 4 to 60 is essentially the same as the current system.
I've always thought it was to differentiate the new scores from the old.
As an example, a 32 on the 'old' exam and a 32 on your theoretical scoring system would be very different scores. Thus, admissions would have to look at the date taken for all the scores. As it is now when someone sees a score in the 500s you know they know it is the 'new' MCAT and when someone sees a score in the 20s or 30s they know it is the 'old' MCAT.
472 to 528 is just an arbitrary scale that the AAMC proposed. The real importance rests in percentiles and percentile comparisons. That's why LizzyM scores can still be used effectively regardless of the scale the AAMC uses.
tests have random scoring so adcoms don't subconsciously confuse them with a) other tests (SAT, LSAT, pre-2015 MCAT etc.) and b) a typical 10 point scale (that's why pretty much no standardized test is 0-100. a 70 is a "C-" on a 10 point scale, not a good score, but on a given standardized test maybe a 70 is actually a really good score)
as someone mentioned previously, percentiles are what's important