Bug in Computer Analysis?

In the computer analysis of this position I encountered in a recent game, it says the score after 18. Qb1 is (-Mat04), which according to my understanding means that I have a forced checkmate in at most 4 moves, and it shows the continuation starting with Nf3+. Then after I play Bxh3, it informs me that that is a blunder and that I should have instead played Nxf2 (a different move than it said leads to mate in 4 after my opponents move). I downloaded the PGN and ran the analysis on my friend's Chessmaster 10 program and it looks like there is no forced mate there at all, so unless I am missing something, it appears to be a bug with the chess.com computer analysis.

I don't think its a mate but it is certainly a blunder.  You definately lose a piece.

18.Qb1? NxNd2 19.BxNd2 BxBd3 with the next move exchanging your bishop for your opponents Rook.  You are down 1 rook!

Well the best move after 18. Qb1 is 18...Nxf2, but the point is that the analysis says it's (-Mat04) which appears to be incorrect. If that's the case it makes me a little leery about trusting the computer analysis from this site.

It was hidden in the analysis give by the site machine (posted above), not something that I found.

But it loses to nxf3. This sites engine has known to be buggy.

Yes, you're right the site analysis is at fault. Pity though, that's a beautiful combo in the bin. I've just run the position through DR4 and it reckons Nxf2 is 12+ points in Black's favour.

Thanks guys, I saw the computer's 18...Nf3+ continuation in the analysis and I was going crazy trying to figure out how it was a forced mate since it looked like 19. Nxf3 stops it. I guess it just goes to show that you shouldn't assume that the computer is always right!