Even if the TCO for Windows is "usually" lower (to avoid the precise number), there is a hidden implication: Since the TCO of Windows is usually lower, that's what you should use. But (there's always a "but" when it comes to statistical analyses) that means that there are at least some cases where TCO is higher for Windows.
The question then, is: What do the cases where the TCO of Windows is higher have in common? Is it based on the industry? The computer-skills level of the staff? The specific programs used? The network effect of dealing with customers/suppliers? The inertia and cost to switch?
Considering the fact that Windows already has a 95% share of the business desktop market, the fact that they can only show lower TCO in 95% of the cases is telling. It means that even with all the items mentioned in the previous paragraph, they're only breaking even on the cost to stay or switch.