As of now, returning an empty hypothesis will always give perfect purity and coverage. This shouldn't be the case.