Distillation works best in very exact sciences, such as physics and mathematics. If you rely on distillation for an inexact science, you will do best at capturing its exact parts. You will be left with a systematic bias, and knowledge gap, regarding its inexact parts.
I’d like to quote this argument from here: