David, it was late yesterday, too late for me, and I even forgot about the unusually high aspect ratio (7.1) of the CAP 21 even though you mentioned it. I think that may contribute to the snap behavior. The Z-50 has similar behavior with only 5.9 aspect ratio, but it has conventional NACA 00xx airfoils. After some experimenting I'd say it should be possible to make the CAP less snappy with a suitable airfoil. It shouldn't lose much lift when stalling, what the NACA 00xx do, though, and the E 169 does as well. The other modern aerobatic airfoils with very forward max thickness (less than 20%) might do the trick. E 472, 473, 474, 475, and 479 might do as well but they are suitable only for bigger models and have much increase of drag with AOA. (By the way, did you read that they tested the type in a spin wind tunnel?)

The CAP 21 has a quite big stab and huge elevator and long tail moment arm so you may pull it to a stall AOA very easily. Due to the high aspect ratio it will stall even at 17 degrees. Besides, it has a pronounced mid-wing to tip stall (giving better snaps) due to the special wing planform. Not as much as modern aerobats but enough to require a special airfoil (my guess, I'm not an expert). Like me, you may find it interesting to compare the planforms of various aerobats (Z-526AFS, Z-50LS, CAP 21, Su-26M, EA-300S, Edge 540) with comparison to elliptical planform.

My simulator is good but still a rather simplified rendering of a model and stall aerodynamics. I tried the behavior with a 1:4 scale model of the CAP 21 and found it snaps and spins like crazy with only 25 degrees elevator (30 degrees rudder aren't even needed to snap). An airfoil which doesn't lose lift in stall calms things down.

There's no parameter for the stab position, but moving the wing 2" up turned the substantial belly pitch into a bit canopy pitch (1.5 degrees dihedral unchanged). I find the belly pitch not really bothersome and would mix it out.

By the way, 1:4 scale (2.02 m / 79.5") really seems too small. It gives 890 sqin wing area, and with 9.5 lbs weight the model is somehow similar to the Brushfire. You might deviate from exact scale by clipping the wings, as you mentioned, what would reduce the snappyness and make it easier to build a strong but lightweight wing.

A 1:3.5 version, exactly scale (2.31 m / 91"), at 14 lbs (6.35 kg) is noticeably less snappy, and if you could build it to 11 lbs in that scale it would be just perfect! By the way, the three-view of the original seems to show about 1 degree incidence for the wing. Even less than 1 degree removes the belly pitch but requires a bit more down elevator inverted. Another choice.

I used 0.25 degrees wing incidence and 0 degrees stab incidence (4% static margin), giving neutral trim upright and nearly neutral inverted. 3% static margin might make it neutral both upright and inverted. But how about giving both wing and stab 1 degree incidence. That makes the plane look better in the air, again both upright and inverted.