I've wondered about your question myself. It is common these days to say, for example, that communism does not work because "look what happened to the Soviet Union." Or that lack of regulation of financial markets does not work because "look what happened to our economy." But of course, someone who wants to defend communism or capitalism can put forward an "ad-hoc hypothesis" that the reason the Soviet Union "failed" is not because of Marxist theory but because of corruption (or bad weather or alcoholism...) and that the reason for failure of the financial markets is corruption (or poor people not paying their loans or public anxiety...). I like to think of political theories as scientific theories--it is common to get contrary data, and typical to construct an ad-hoc hypothesis to "explain away" that data. But that's not the end of the inquiry. The next step is to go on and see if the ad-hoc hypothesis has any data in support of it, or whether it is simply embraced in order to save a favored theory. Scientific theories and political theories should not be accepted or rejected based on a few data points, but, rather, they should be evaluated in terms of a range of data and theory modifications.
Perhaps you are also/instead asking specifically about recommendations for human behavior, and wondering whether they should take into account likely human failings? Philosophers often say "ought implies can." That is, any recommendations for human behavior should be realistic recommendations for us qua fallible human beings.