I use centipawn loss average from Stockfish analysis to determine how accurately I played.
I've stopped caring about chess.com accuracy after I read somewhere the chess.com accuracy gets tweaked depending on the elo bracket you are in. I dont think it's very relevant.
The chess.com accuracy figure is just a chess.com thing - in analysis, the standard metric is centipawn loss - to either evaluate moves individually, or a sequence of moves.
I favor whatever degree of accuracy the game I'm winning has.