You are equating the number of right answers to a score and that is not how the scoring algorithm works. It depends upon the difficulty levels of those questions. You can get exactly the same number of questions right and get wildly different scores.

I don't know much about the algorithm PR uses, so I can't really comment on the accuracy or inaccuracy - I just want to make sure you know that it isn't about how many you get right, but about the difficulty level of the questions you're answering.

In general, though, if you are getting consistent results from 2 sources (MGMAT and GMATPrep) and one inconsistent result from another source, then it's reasonable to take the outlier with a grain of salt. I wouldn't worry about it too much.