I'll just comment that there's no easy way to flag data as reliable enough for some purposes, while unreliable for others.
A FBW system using alpha as part of the control laws is rather more sensitive to AOA accuracy than a conventional type 9which frankly hardly cares until you get to low speeds where the AOA should be doing stuff). i *suspect* - and I confess its nothing more than a semi-educated guess - that one of the reasons for declaring AOA invalid at low airspeed was that the AOA vane could not be stated to provide the required accuracy at such low speeds, and that this accuracy was specified in the context of normal law operations. Gross AOA, good enough to fire the shaker, wouldn';t be good enough to drive the FBW laws.
Now, you could say "ah, but you can't be in normal law at such low speeds" - and normally that'd be correct. But the design was probably trying to cover all bases, but with the assumption that defending the normal law scenario against erroneous and dangerous AOA data was the more important thing to do.
Now, with hindsight, we might try to provide for discarding the AOA data for the control loop closure, while retaining it for the warning. But that's also complexity - and the golden rule in designing complex systems is to not make them any more complex than they need be. Judging the point of "need be" is the trick, of course.