Why is a physics professor surprised by this? You will get exactly the same "step" plot directly from the binomial distribution if you assume independent trials and a linear dependency between the SAT band and the probability of getting an A in a course.
If someone's probability of getting an A in a single course is 0.7 or above, the probability of getting at least 8 As out of 16 courses rapidly converges to 1, similar result holds for those with probability below 0.3.
Nothing new has been shown here - can't believe they actually wrote a paper.