The problem with ratemyprofessor (and virtually any 3rd-party "review" platform) is the potential for participation bias. Students only participate when they 1. Know about the platform and 2. Have strong enough opinions that they feel a need to share. This means that the ratings are biased in favor of the exceptionally good and exceptionally bad. (throw in a little human psychology, and the bias moves towards bad reviews appearing proportionally more often than they should).

The advantage of having the data come from Yale is that they can mandate that every faculty member must take class time to conduct the student evaluation. This ensures that [nearly] every student's experience in every class will be considered in the final results, which gives much more legitimacy to the results.