It seems to me that the disagreement here is because you’re looking at different parts of the problem. It might well be said that you can’t have a well-calibrated prior for an event that never happened before, if that entails that you actually don’t know anything about it (and that might be what you’re thinking of). On the other hand, you should be able to assign a probability for any event, even if the number mostly represents your ignorance.
It seems to me that the disagreement here is because you’re looking at different parts of the problem. It might well be said that you can’t have a well-calibrated prior for an event that never happened before, if that entails that you actually don’t know anything about it (and that might be what you’re thinking of). On the other hand, you should be able to assign a probability for any event, even if the number mostly represents your ignorance.