That's a common misinterpretation of "correlation does not imply causation". It'd be more accurate to say "correlation implies causation but the direction of causation must be derived separately." If there's a real correlation between A and B (i.e. the correlation isn't a coincidence), that must mean that either A causes B, B causes A, or a third factor causes both A and B.
In context, since we know that femaleness is (barring certain rare disorders) caused by the sex chromosome of the sperm that reaches the egg, which is essentially random, we can be fairly sure that it's not caused by school grades or anything that would affect school grades, which does allow us to conclude with a reasonable degree of confidence that femaleness (indirectly, of course) causes higher grades.
> If there's a real correlation between A and B (i.e. the correlation isn't a coincidence), that must mean that either A causes B, B causes A, or a third factor causes both A and B.
Pirates being correlated with global warming isn't a coincidence; pirates decreasing and global warming are both (indirectly) caused by the Industrial Revolution.
It's true that you can never be completely sure that any correlation isn't a coincidence, just like you can't be sure that the sun will rise tomorrow. You can, however, be sure beyond a reasonable doubt with enough of a sample size and a strong enough correlation.
Out of curiosity, do you personally believe it's likely (say, >5% chance) that it's a coincidence that girls have better grades than boys?
My interpretation was that he suspects that the higher performing female students were the ones that were less likely to drop out due to the negatives that come from the male-domination of the subject. So rather than the professor's assertion that "females are the highest scorers", he is suggesting that "the females that don't leave are the highest scoring females".