By definition, passive movements (those produced by the therapist instead of the patient) involve no CNS activity, and therefore lack the ability to affect the CNS. Moving a joint through a ROM without the use of the patient's neuromuscular system seems like an excellent way to postpone any therapeutic value that might be obtained in the therapist/patient interaction. In the case of partial paralysis, the patient engages as much of the kinetic chain as possible and obtains adaptation around the completely paralyzed portions of the KC. As nervous function returns, the portions peripheral to the damage continue to assist the damaged areas as their function returns -- IF the gross motor pathway that uses the whole KC has been trained. PT's problem is that it thinks about muscles instead of movement patterns.
So, you get the patient to perform something that looks as much like a squat/deadlift/press as you can. This requires a background in coaching these movements with unimpaired subjects.