I'm always grimly amused at how some wonderfully stable computer models (e.g. the ones the Club of Rome used in Limit to Growth) can be dead wrong despite being based on sound observations. But, they are still scientific theories in Popper's sense. (They just happen to have been falsified, and therefore shown to be wrong scientific theories.)
f.o.g. seems to not like my sociological point about scientific theories (whether good or bad) also becoming religious dogmas for secularists, but wants to deflect it by claiming the bad theories of the anthopogenic global warming crowd aren't scientific theories.
Better is using the models to accurately model recent events, because that data is pretty good. The "Pinatubo test" has been a benchmark -- models had better get the cooling right if they're to be expected to get the warming right, too.
Though it's old, this is still a good article to read: Forcings and Chaos in Global Climate Change
Here are a couple of others:
Greenhouse Gas Influence on Northern Hemisphere Winter Climate Trends
The Sun vs. the Volcano: Drivers of Regional Climate Change (I think you'll appreciate this one)