You are correct that all hazard analysis includes consideration of both the probability of the event and the severity of the consequence arising from the event. Low consequence events are generally permitted higher probability of occurrence while still being regarded as acceptable. But then you miss the essential point of the methodology - because for each event under analysis, it is always the worst case consequence that is used to determine the hazard rating (risk).
In this case the event in question is a drone/aircraft midair collision. To determine the hazard rating, and thus whether the risk of this event is acceptable, one considers the probability of such a collision and the worst case outcome (consequence) IF that event does happen.
So - I was discussing severity of consequence, which is always worst case, using a combination of knowledge of terminal ballistics and the educated guesses that you mention. That led to my conclusions that drone/aircraft impacts cannot be ruled as low risk on the basis of low consequence. But that does not lead to paralysis of anything, since all that is required for the risk to be acceptable is to be able to reduce the probability of the event to an acceptably low level.
In the case of a midair collision with a slow-moving flying object, impact with the airframe, windshield or engines are all quite common, and so we take no credit for a significant reduction in probability of any of those scenarios as a subset of midair collisions, and so we have to reduce the probability of midair collision as a whole. Which, in turn, lead to my observation that the increased use of drones, based on the huge increase in ownership and numerous well-documented examples of drones being flown in the vicinity of aircraft flight paths, must, by definition, be increasing the probability of a collision. That trend is inconsistent with the requirement to keep the probability of a collision low to offset the high event consequence.