Few of OOXML's Flaws Have Been Addressed
As well as with the original article. First thing - you can't really say "few flaws have been fixed" when the original article (and the post blurb) specifically say that no fixed flaws where actually found in the testing sample.
On the other hand, the statistics used by Rob Weir are shoddy according to my local statistics semi-expert (my girlfriend who finished 2nd year BA stats A. with a perfect 100 score). Specifically his sample is incredibly small: 25 random pages out of a random selection of 200 pages out of 5220 pages of the original standard document, out of 6045 pages actually in the original document (not the amended document), of which he doesn't know how many defects where actually reported against each page (we know how many were reported totally, but we don't know what is their percentage in the first sampling or subsequent sampling), and as Rob Weir found new defects that were not reported to Microsoft in time for the BRM, he has no idea what is the actual density of (pre-BRM) reported defects in the total "defect population" (defects discovered before BRM, after BRM and defects that are yet undiscovered).
As such a confidence interval of 1.5% +-3% (i.e. at worst 4.5%, which is not what the post reports) seems highly suspect. To clarify for non-statistics students, a confidence interval of 1.5% +-3% in a result of 0 hits out of a random sample, means that Rob Weir is at worst 95.5% confident and at best 100% confident that there were no defects addressed by Microsoft.
This is awfully presumptuous, even if its Microsoft that we are talking about.