I don't really understand the idea that tests aren't a reasonably good reflection of actual intelligence. They can be studied for, sure, but everyone can study for them, and the more intelligent you are, the better you will be able to study, or to turn that study into results. It's of course not going to be completely accurate (if you fluke a test, or if someone intelligent just doesn't study), but is anyone seriously contending that some other arbitrary criteria (entrance based on interview questions, for instance) is? There are certainly problems with standardised testing, but basing university admission on academic achievement would certainly seem to be a reasonable way of reflecting intelligence.
They're an indication of how good you are at tests, which depends on how intelligent you are to a huge extent, but also how good your schooling was (which depends on where you live, whether you went to public school, and so on) and how well you've prepared for the test, which depends on your ability and dedication but also on the resources and support you've had for it. The idea is to make tests which are minimally reflective of these things (usually by asking people to apply knowledge - any idiot can learn that Franz Ferdinand was shot in Sarajevo or that secret alliances helped to lead to war, but you can't answer a question to the tune of 'was the shooting more important that the alliances?' purely on the back of what you've been taught, because there are far too many possible questions like that to prepare for them all. Interviews, which require intelligent responses to be thought up very quickly and often ask questions for which there can be no preparation, also seperate, as I've said before, great candidates from very good candidates who look great in exams.