What grinds my gears with all the people (whether Denuvo officials or elsewhere) that claim that it has no effect on performance: they only focus on average FPS. Never a consideration for FPS lows or FPS time spent on frames that took more than N milliseconds. Definitely not any look at loading times.

I'm willing to believe a good implementation of Denuvo has a negligible impact on average FPS. I think every time I saw anyone test loading times though, it had a clear and consistent negative impact. I've never seen anyone check FPS lows (or similar) but with the way Denuvo works I expect it's similar.

Performance is more than average framerate and they hide behind a veil of pretending that it is the totality of all performance metrics.