Your other ratings are Elo, and based on how you play against others. Chess mentor is based on arbitrary problem ratings set on problems, which vary drastically (e.g. some have a large number of "alternative correct" answers, some don't bother to set them, even when they acknowledge that they work).
They're not really comparible. Then again, my ratings vary wildly anyway I'm 1150 blitz, 1600+ correspondence, and 1850 Chess Mentor.
If you have a look at my profile, you will see the huge difference between my Chess Mentor rating and all my other ratings. So is there a rating flaw?