Re-reading the article, I think that I know where you got the 90% number from, but I think that you are wrong.
Firstly, as you note in another post, it's 89%, not 90% (bias showing here?)
However, more importantly, it's 89% of all the "positive" results, not 89% of those used at trial. In the 11% where there was a false match, the likelyhood of actual guilt is lower, so the amount of other evidence would be much lower, thus it is very reasonable to assume that the 11% of false matches are over-represented in the 268 cases where the hair evidence was used at trial. Thus, your estimate of 27 or 28 is likely to be very low.