These numbers have a habit of being misleading. To properly understand the numbers you need to think about:
1. Is there a seasonal aspect to computer sales? Vista and 8 were not released in the same month.
2. Is there an overall trend away from buying new laptops? We are in a recession, after all.
3. Are there any seasonal glitches to look out for? For example, is it possible that somebody buying a machine on Dec 1st might not power it on until Dec 25th for example, and thus the number might not register on the website?
4. Are the users of the website themselves a bad sample? For example, if you did the analysis on Slashdot I'm sure you'd find that a huge proportion of everyone everywhere are definitely using Linux.
5. Has the number of computers on the web changed since Vista's launch (for example, is the number of people using a browser on a phone/tablet different now to in 2006?) 2.2% of 1000 is 22, but 1.2% of a million is 12000.
6. Has the number of people visiting their website, or the demographic of their sample changed? Is it still (or rather, was it ever) an unbiased sample of the population? - for example, are business users using IE6 to browse IBM documentation and NASDAQ tickers but who don't waste time reading tech enthusiast blogs blog as equally represented as tech enthusiasts who might browse blogs all day on their iPad, iPhone and Macbook?
A better comparison would be to compare Windows 8 sales as a proportion of total Windows sales compared with Vista sales over the same period and to compare what percentage of desktop/laptop PCs are still distributed with Windows.
Sadly I suspect getting better quality numbers would be harder, require a better understanding of basic statistics and would probably also make for a less sensationalist headline.