Right, I read this - the data she is using is a bit skewed. First of all, her numbers are only reflecting allopathic physicians. She fails to take into account that many DO programs pride themselves on overlooking certain low scores to a degree if the candidate displays personality traits favorable to being good doctor.
With that being said, the MCAT is needed in order to see how a person may tend to perform on the USMLE. Even if a 'personality test' were implemented, eventually one would have to take step 1, 2, and 3; and once again the competition is present.