I agree with perrin
on this one. I had to take a Perl coding test once and I was completely comfortable with all of the tasks. However, as I started working it seemed the test database wasn't working. So I started debugging the problem by writing some test scripts. They came back to check on me and I hadn't really progressed on the task at hand.
Turns out the database was broken and they wanted to see if I would ask for help. Well, in my office I'm usually they guy to figure out these types of problems so I don't usually have someone else to ask.
I guess my point is it's really hard to have a test that accurately measures everyone on all criteria.