99,5 with many mistakes, why?

I think they recently increased the numbers, to make everyone feel better about themselves.

I'm regularly scoring 95%+ on my blitz games where I know I'm making heaps of blunders and inaccuracies.

As long as we don't know how the accuracy is calculated, it's quite meaningless.

See this one i got less accuracy just because I had 30 best moves and he got 31, and what about blunders he made?? Is this some kinda bug?

While I have no idea how accuracy is calculated, I suspect your higher average difference also factors into calculating a lower percentage. 

Well, in a 10000 moves game, 0.5 % mistakes means 50 mistakes. 

Oh! You analysed it? Thanks for your reply. But still that's a 0.08 difference(only) and Why should he get a better accuracy with 4 blunders more than me?

You can find those stats within the "Details" tab when looking at the analysis of your game. There are other interesting stats presented there too, such as accuracy per piece.

Some of your opponent's blunders resulted in massive differences from the best move's evaluation. Who knows how Chess.com even calculates the average difference then? Perhaps it caps individual differences at some figure? Quite often, if you let the engine run to a deeper depth, large evaluations turn into forced mates, which further complicates an average difference calculation.

I quickly learned to ignore Chess.com's analysis due to it handicapping Stockfish with a very low ply count of 18. When I want to analyze critical moments in a game, I often get quite different evaluations by letting the engine achieve a higher ply count in self-analysis mode.

Apparently you know exactly what this accuracy rating is supposed to mean. Please share your knowledge!