The AI coupling system relied on a partial cutout of the aperture ring to convey the maximum lens aperture to the meter.
What it did in reality was change the aperture setting value that was conveyed to the meter to a "standardized" value that was uniform for all lenses.
All AI lenses had the meter coupling shoe that all F2 metering prisms require, so all AI lenses would work with all F2 metering prisms.
so what you are saying is it could still easily be that the guy's meter is weird. Absent more careful testing, I'd at least advise him to assume as much until proven otherwise. External meters are more accurate anyway.