It's hard bordering on impossible to make a single test that measures most of the college-bound high school graduate population and can also make non-arbitrary

It's hard bordering on impossible to make a single test that measures most of the college-bound high school graduate population and can also make non-arbitrary distinctions among top-1% students. Lopping off the top of the distribution looks to me like a recognition that pre-recentering scores in the top 100 points or so were essentially measuring the luck of top-performing students on which questions they drew and how well their educated guesses landed.