When it's consistently in favour of Windows 10, then it's not the statistical error..
The author's conclusion doesn't match your Dunning-Kruger interpretation of his data.
As evidenced by you using a boot up time graph to prove your point on minimum frame rates and microstutter.