Obviously SAT scores were a bad place to look for differences in education between schools.
What's interesting is that analysis failed to reveal a difference in the number of AP tests taken, and passed, despite the fact that the exam schools offer so many more AP courses. It seems that instead of taking more tests, exam school students are simply selecting a similar number from a larger domain. I don't see that the study attempted to make any distinction between taking and passing a Calc B/C test rather than a Calc A/B test though, which seems like a flaw to me.
On the other hand, is it reasonable to expect that it's no better to be in a class where the pace is a little too fast than a little too slow?
If we want to know the value of exam schools, shouldn't we focus on the students who are more typical of exam schools? The students who got in (not borderline) but didn't go because of geography, etc?