Partly it's out of date but also it's skewed by measuring pages on secure servers. Those are largely used for e-commerce and a very high proportion of them are in English (or at least were in English as of 2001, it's hard to say how that might have changed by now).
If I remember right, the OECD report in question's look at links to secure servers (https, really) -- outsourced to Netcraft, who did the actual work -- was an explicit attempt to measure e-commerce activity, not general Web activity (which is being referred to here as "Internet" activity, apparently). I think that right next to it was a comparative figure for http (non-secure) links, where English was a bit lower.
Some statistical work was done for UNESCO here last year that indicates that there is vastly less language diversity on the internet than is routinely claimed.
And I'm talking about whole orders of magnitude of difference, here, not just a few percentage points.
Elijah, that sounds quite interesting -- can you describe the methodology? cheers Bram