I think I've solved the cue dot puzzle. As discussed above, it's the azimuth in the frame of reference of the jet, where the planes wings are considered horizontal.

I've added some fun new features to the sim to visualize and measure this.

The cyan (light blue) plane is the frame of reference of the jet (i.e. it's a plane parallel with the wings - in the abstract sense where the wings are perfectly horizontal)

The yellow plane (seen edge on here) is the horizontal plane.

Azimuth is the angle between forward and the camera's vector in whichever plane of reference. The Az number shown on screen is in the horizontal plane. The Cue angle is in the blue plane. So to calculate the angles we need to project the the view vector (the line from the camera through the white dot) one the respective plane, and measure the angle between that and forward in that plane (the projection of the plane's axis onto the plane.

These calculations, verified graphically like this, allow us to calculate a plot of the cue angle, and compare it with Az. This shows up if you select Show/Hide cue data:

It's the same. It started out as a good match, and got better when I changed AoA from 3° to 5°, which was something I'd suggested doing earlier as it made for a nicer fir with the roll curves. So that kind of validates the 5°, which, as someone suggested before gives a way of calculating the AoA - although not a simple one, as you have to draw the graph accounting for bank - and AoA may well change.

However I feel this should put to bed any notion of the Cue angle being a problem, as it's exactly consistent with the rest of the data. This is taking data scraped from the screen, and showing a perfect mathematical relationship. It even matches the little kinks during bank changes.

Why the divergence at the end? It's just the way the angles work out - combined with the various bank angle changes. Difficult to explain, but you can watch it happen with the arrows in the sim. Here's Az =0, the Cue angle is lagging behind, but it makes perfect sense projecting the view vector onto the relevant planes.

Then a little later, the cue dot is centered on the screen, but Az, as you can roughly tell from the 5° ring, is about 5° ahead