The funny thing is that there is no technical reason at all for compositing to have worse latency, even for games.
Think about the actual operations that are involved. You certainly never want to render directly into the front-buffer (you'd end up scanning out partially rendered scenes). So you render to a back buffer. Which you then blit to the front buffer (assuming you're in windowed mode; in full-screen mode the compositor goes out of the way anyway and lets you just flip).
The only difference between the various modes of operation is who does that final blit. In plain X, it's the X server. In Wayland, it's the Wayland compositor. In X with a compositor, it's the compositor.
Now granted, some compositors might be silly and re-composite the whole screen each frame, but clearly that can be avoided in most cases.
Depending on the setup, there can also be some issues with the scheduling of the game's jobs vs. the compositor's jobs on the GPU. Valve are working on this currently for VR, since the problem is much more noticeable there -- clearly it can be fixed one way or another (on Radeon GPUs you could do a blit via asynchronous compute if need be, for example), but note that compositing actually doesn't change this issue (since the X server's jobs also need to compete with the game's jobs for scheduling time).
So if compositing has worse latency, it's because nobody has cared enough to polish the critical paths. Conversely, compositing clearly does have advantages in overall image quality. So why not fix the (entirely fixable) technical problems with compositing?