Rats are, in some ways, very different from humans too. But if it's impossible to run human trials, or there are severe deficiencies in the way we are able to execute them, studying rats is still the best argument.
If there is a convincing empirical economic study that widely contradicts a theoretical consensus---yes, that evidence should be preferred. But that is not often the case, and otherwise the theoretical argument is simply the best argument there is.
What drives me bonkers, though, is the people who dismiss the math and its conclusions as being 'meaningless', but then what they substitute is not superior empirical arguments of their own, but their own emotional sentiments and undisciplined intuitions (typically decorated by cherry-picked historical observations).