It’s not surprising about Princeton because the ranking includes quantity of publications as well as quality, and quantity of publications is going to be directly correlated to how large your graduate research programs are.
Princeton has only 2600 grad students, total. Looking at that list, I see that Harvard has 14,000 grad students, Stanford has 11,000, Johns Hopkins has 15,000, UWashington has 14,000, Berkeley has 10,000, UMichigan has 15,000, UPenn has 11,000 and so on.
I suspect that the correlation is pretty direct between number of grad students and number of publications, but I have no direct proof of this.