I bet it's difficult to normalize because performance depends on so many things. CPU cycles, but also other factors like RAM speed, swap usage, CPU cache sizes and speeds, background tasks, GPU and GPU drivers, OS, display servers (X11 vs Wayland on Linux), window managers, available fonts, and probably a ton of other things are other factors that will play a role.
Something so that more megahertz/cores doesn’t throw off the number.
Whatever the reason, it is nice to see it improving. I think it would be interesting to also see a median figure in addition to a 95% figure.