swa66 - 9:01 am on Jan 10, 2013 (gmt 0)

Tests: the score doesn't tell it all.
First of all such tests usually are not 100% complete in coverage of what's tested. So since the tests are known developers can focus on just covering what is tested and leave all the rest untouched. That way you can score better on tests without actually being good.

Next missing out on a feature doesn't always have the same impact. There are e.g. a number of html5 input things that are marginally if at all supported by most browsers. But missing out on this means the browser falls back on what it does understand and that's good enough to get the job done - it's annoying, but no disaster. But e.g. getting a box model wrong well that is a disaster at every turn or twist - <sarcasm> not that IE has ever done that of course </sarcasm> ...