My view is that comparative anti-virus tests are a bit like supporting your favourite sports team. Just because they won the last 6 times, it doesn't mean there's any guarantee they'll win this time. But they are a bit more likely to win than the team who lost the last 6 times.
Same with virus tests - just because an AV product wins lots of tests, it doesn't mean it will stop the malware on that website you look at tomorrow. But it is perhaps just a bit more likely than the AV which lost the last few tests.