Thanks for the input guys! I am interested in both "best practice" AND what is good for this specific model.
ChuckAuger is right that the GeeBee will NOT need 3D throws... as a matter of fact, only 5/8" in both directions. However, because they have you use the inner-most hole in the servo arms (5/16" from servo-center), and a long horn on the elevator, the servo-arm will go through a large arc to provide this (as shown in the diagram below). This should aggravate the problem... as does the fact the pushrod ends don't start at the same point at neutral (as in MinnFlyer's diagram).
Here is an illustration of what I see to be the problem:
The black pushrod stays pretty much parallel to the stab, where the red pushrod's angle increases the further off center you go. The red pushrod will not push the horn as far as the black pushrod, because of this geometry. Playing around with the drawing program, which lets me rotate and move lines and groups of objects around, seems to confirm this. In the bottom section, I would have to increase the length of the red pushrod to deflect the surface as much as the black pushrod does... with the same degree of servo movement (shown at 45 degrees in the diagram).
This specific model has been reported to have a tendency to "tuck" in high-G maneuvers. Could this be caused by this slight difference in elevator travel?
Any furthur thoughts?