Hi Brian!
It is very difficult with my bad English explain all what I want, but I try it.
The camera lenses are centered, the eye lens system is decentered.
See image decentered.jpg
In other words: when our eyes see parallel we have a certain convergence. So the convergence is built-in in your eyes.
See image: decentered01. jpg
The parallel cameras dont have this "built-in" convergence and for this reason we have problems at the edges on our anaglyphs.
If you see:absent.jpg, you will understand that we have constantly need a little of FOV and film. This problem was resolved by the Realist, shifting the film. But I think this isnt enough, because is absent the real convergence ( the "built-in" ).
The big goal made by Fuji W1 is: the Fuji made the real "built-in" convergence and made too the "film-shift" too.
Remember? The 2D mode has 35 mm focal at the Fuji, but the 3D mode is only 39 mm. What it means?
It means that the Fuji has a possibility to shift the film in a certain limits.
Fuji.jpg: see the limit of the fil shif on the left lens, what isnt converged,
On the right lens the Fuji has the "built in" convergence too, and the same possibility of the film shift.
I agree that the convergence isnt the best way because this produces keystone effect, but since the convergence is about 2 degrees, the solution is acceptable.
One of my friends (Antonello Satta)  made the theoretical ideal stereo camera. It is based on "shift lenses".
See: ideal stereocamera. jpg
So the conclusion: if we have 2 cameras the suggestion is this:
Use a convergence of 1-2 degrees ( enough on only one camera ) and use a wide angle lenses, but at the final crop use less image size than the original. ( exactly what done the Fuji )

Imre