(Disclaimer: I am a scientist, so this isn't anecdotal)
Mod my parent up. This is -precisely- true. The idea that we test a hypothesis and refine it based on experimental outcomes is utter BS. In all but the most -basic- of processes, there simply is no way to account for all possible results of testing a system; this is further compounded in my field, where an 'in vitro' experiment may yield different results than one 'in vivo'. To make matters worse, those 'in vitro' experiments may in fact yield different results in /different cell types/, given all other conditions being precisely the same.
Without going into too much detail, the real nature of science is that we already have a fairly good idea of what we -want- to happen before we begin testing. I may have a vague theory about the experiments will come out, but more often than not we end up writing the theory to fit the facts around the time the data is published, in such a way that it fits the data we've collected, even though that final theory may not have any relation to the initial expectations.
Some of this is also attributable to the funding system (at least in the US). Submission of a grant (money to do experiments) requires that you already have (preliminary) information, and a fairly tight and detailed set of theories to explain how what you propose to do will result in a conclusion, as well as what those conclusions will be. Essentially, you need to present some data in order to get funding to obtain data. Give too much preliminary research, and you won't have enough theory and interesting suggestions to get funding, but if you don't have enough research done (how you do this without money is a nice conundrum), you won't get funding.
In practice, this often means researchers with no active funding will dust off old unpublished work, and write theories around it, in order to talk the NIH into paying out money so real work can get done (since once you're funded, you can really do whatever research you want -- especially if we're not talking about a renewable grant).
So it's really a messed up system all around, but the scientific method as you know it has virtually no role in it either way.