He seems to have better evidence than the original claim did, though. He quotes statcounter’s own method of producing their figures, and what they actually mean.
He points out a very odd anomaly with OSX vs macOS numbers too that highlight the problem even more.
Excellent points. I hadn’t thought of that.