How inflated are ratings?

Uh, I'm not playing competitively, but I'm pretty sure my rating wouldn't be as high as chess mentor is giving me.  How inflated is this rating?  Or does it not mean anything?

By the way, I'm loving the program.  I think I'm learning loads.

Chess Mentor gave me a rating of 2100+ when I was a premium member, about 350-400 points higher than my USCF rating.

That sounds about right for me too--if I ever got a USCF rating, I'd expect it to be around 300-400 below what chess mentor is giving me.

I would agree around 300 points different, however it also depends on how well you play under pressure. I find OTB tournament games are very stressful and mentally taxing compared to online chess where I tend to not care whether I win or lose.

I think in OTB games I tend to out play myself and go into a passive shell.

Chess Mentor ratings are only that, Chess Mentor ratings, not online chess ratings or OTB ratings.

Chess mentor and USCF are simply two different rating systems, the former merely attempting to resemble otb; the guys who made chess mentor ratings obviously had to estimate the strength required to solve any of the lessons which is going to cause a lot of imprecision. Besides, it's hard to really make a cutoff sometimes: sometimes a 2000 level player (otb) may not understand something about pawn structure, but a 1600 does but is bad at tactics, and so despite understanding an "advanced" lesson it misrepresents his strength.

Rating is a perfect measurement of results; it is a highly imperfect method of measuring strength and only takes into account wins and losses above anything else. Alas, that's the best method we have.

On ChessMentor, you know there is a combination to be played, and that it is worth taking your time to find it. In a real chess game, maybe only 1/3 of the moves are blunders. If you spend 5 minutes searching for blunders on every move, you will lose on time. That is probably the reason your rating on ChessMentor is higher than your actual rating. The difference between an 1800 player and a 1500 player could easily be that the 1800 does not make nearly as many blunders, and catches most of the opponent's blunders. You can play like that as well in correspondence chess, which does not reveal your true strength.

Chess Mentor ratings are definitely inflated but by how much I think varies by the player. For instance, my C.M. rating is around 2350, however, in a short time control game I find that I play around 1800. But when I play more reasonable time controls, (like 40/120 G/1, or G/120) I can play around 2100-2200. 

The problem with trying to get a rating from puzzles is that the idea is already present in the position. What differentiates the Master from amateur is the ability to see the possibility of different ideas and putting the pieces in the right position to maximize their potential. I think of it like two people driving automobiles. The amateur is driving an automatic transmission, the Master, a manual shift. Yes, they are certainly both driving, but the Master is doing things at a more complex level.

Actually, correspondence chess does reveal your true strength... for correspondence chess.  You can be better than the average player in correspondence chess, but worse than the average in blitz, or vice-versa.  Neither game can in any sense be "easier" for everyone.  They each play to people's different strengths.