That seems to be changing the definition to fit your claim…
Going through A-C in the graduate student directories for mention of undergraduate school:
Stanford ( https://mathematics.stanford.edu/people/graduate-students/ ): 1 Michigan State, most no information.
MIT ( http://math.mit.edu/directory/profile.php?pid=1641 ): 1 UCB, 1 MIT, most no information.
Harvard ( http://www.math.harvard.edu/people/graduate.html ): 1 Cambridge, most no information.
Princeton ( https://www.math.princeton.edu/people/graduate-students ): no information.
Chicago ( http://math.uchicago.edu/people/grad-students/ ): 1 Yale, 1 Princeton, 1 Fudan, 1 Virginia, most no information.
Brown ( https://www.brown.edu/academics/math/graduate-students ): 1 UBC, 1 Colby, 3 no information.
Of the mentions, there are 3 US private universities, 3 US public universities (1 non-flagship), 1 US private LAC, and 3 non-US public universities.
Sample size of undergraduate schools from public information is too small and does not support as strong a conclusion that you are claiming.