I believe the test is intended to see how many new viruses not in the definitions file an AV will detect by heuristics alone, not how well the AV copes if the user doesn't update the definitions file. With malware writers now producing new variants of Trojans and viruses every hour, even the quickest virus definitions update will not keep up with new variants, which is where heuristics comes in. Good heuristics will stand a good chance of catching the new variant, even where the latest definitions might miss it.
EDIT: Typo.