CPU Analysis contradicts itself?!

Yesterday I played a live chess game as Black.

http://www.chess.com/home/computer_analysis.html?id=117118837&game_type=2 

After analysis by the CPU, I noticed that at move 29 it contradicts itself. After 29. Rxc7?? he recommends Nf6 Ne5 Bd6 as the best continuation. It the game itself I played 29... Nf6, which, according to the CPU... is a mistake! Now, he recommends Bf6 Rb7 Ne7 as the best continuation.

I have heard that the 2000 CPU analysis isn't too strong, but how can a computerised system make such a confusing assessment of my game? 

P.S. don't mind the game itself, we were playing casually and under time pressure which resulted in a blunderful game.

yeah I submitted 2 of my games for free analysis and both were useless. I wont use it anymore maybe when I decide to pay for premium (premium analysis supposed to be better)

how much time did you use to analyze every move? in some videogames you can set the time to analyze every move. what i mean is, that computer uses "beans" to analyze best move. more "beans" better move to use

I'm not sure, you can't set the amount of time as far as I know. In any case the game lasted 50 moves and the analysis was done after 10 minutes or so, meaning the CPU took 12 seconds per move on average.

The computer analysis on here can be dodgy at times. It's called my moves a mistake when it's suggestions lead to an obvious forced mate.  It's a good tool at times, but sometimes you're just in a bad position or a "complicated" position.