It depends a lot on the opening and the players.
I have played several 'perfect' games when the opening was one I knew and the opponent made less than perfect moves making my choices obvious. I have also played some games with much worse results.
You can NOT conclude anything from one game or even a few games.
I was a statistician but this is not an area I have expertise with but I suspect that you may need hundreds or thousands of games to make that accuracy thing meaningful and yet may still have a wide error band on it.
I'm a 1400-1500 rated blitz player on chess.com. Looking through my archives, I see that just this month I had a 34 move game, at a 3 2 time control, where I was rated at a 97.4% accuracy. And I know I wasn't cheating.
When I look at my recent daily games (where I'm rated higher, around 1900) I see a 46 move game where I was rated at 99%.
We can't determine a cheater from just an accuracy analysis. Perhaps the player was simply good, or most of the moves were forced, leading to high accuracy. However, for an average player to get 99% accuracy, this clearly is a warning.
There is a better approach to detect cheating. A game can be analysed to see how 'computer-like' the moves are. If a person did not use an engine, then it can be clearly seen that he/she is not cheating. However, if an engine was used, then cheating is evident.