That's because broken theories can be used as frameworks to predict outcomes in coarser grained slicing of systems. Falsification is used when it comes to modelling but as a scientific test or rigor it's still important to pursue it because it finds the limit of your theories.