I wouldn't exactly call VAERS reporting "a scientific process". The disclaimers on the database read like the ones on Slashdot polls.
If it were scientific, the doctor would call you a week after the injection and ask how her health was and if she suffered any major or minor maladies, and the answer whatever it was would go in a database.
Or, heck, all adjacent pairs of medical interventions could go into a database, whether it had anything to do with vaccinations or not. There is much you can do with lots of data; the health care system is designed to squirrel that data away into various different filing cabinets, not look for patterns. Medicine has a long way to go yet to be a data-driven science in the way that, say, advertising is. (Kinda makes you think about our priorities as a culture, eh?)
Even with the lack-of-reporting bias, VAERS still can be used to detect particularly problematic vaccines, simply because some reports do get through, and there's no reason to expect the legitimate ones are suppressed more than the illegitimate ones. The assumption is that _none_ of it is virus-related. So, okay, sample size is smaller and statistics are noisier, but you can still detect obvious trends with careful statistics.