I'm not sure what higher-level point the GP was trying to make[1], but I'm fairly sure they knew macOS doesn't run on mobile. Their implication is that were mobile devices included in the tally they would have contributed significantly to the denominator.

[1] I suspect it is that macOS + iOS would be a better numerator to pit against "Windows on all devices". However the GP having corrected themselves immediately means its not clear what the point is.