Using 2 different cameras in same comp?

If there is to be no depth between the background images, the Theater Stage, then there is no reason at all to have those layers be 3D. You can just layer them. 

If you want a 3D object to appear to have depth then you need to use a standard camera with a standard lens. The default camera fairly closely simulates the normal perspective you get when viewing things from a normal viewing distance.

I think that the biggest source of your misconceptions is that focal length determines perspective. Perspective is only dependant on camera position. There is no way for you to realistically recreate the perspective on any projected image for a viewer on the left side of the screen and a viewer on the right side of the screen at the same time.

If you are projecting on a flat surface just set up the scene to look the way you want it to look and don't worry about anything else.  If you are projecting onto different surfaces you need to set up your layers so the distance from the projector to the surface is at the same ratio as the distance from AE's camera to the layers intended for each surface. Did you follow that?

Let's say that the surface you are projecting on has 3 planes, one 200' from the projector and on the right side of the frame, another 250' from the camera and taking up the center 1/4 of the frame, and the third 3000' from the projector on the left side of the frame. Assume that all of these surfaces are perpendicular to the projector.

If you wanted to be absolutely accurate you would go to the location and set up a camera in the same position as the projector, adjust the focal length to frame up the shot so that it just includes the surfaces, put some reference in the shot that you can measure, like a 6' tall assistant, then take that shot and record the data. 

If you want to just get close just create a default camera (50mm lens), add it to the comp, then double the Z distance from comp center.

Now add 3 solids to the comp - all the same size and position them at the same relative distance from the camera. IOW, the layer on the right would be set to 2000 pixels in Z, the layer in the center would be at 2500, and the one on the left would be at 300 pixels. Now you scale all of the layers to make them the same relative size in the frame. You can use an expression to do that. The last step is to crop each layer so they match the surfaces of the building. This is much easier to do if you have a photo taken from the same position as the camera. Now you have your reference planes set up and you can start creating your scene. 

I hope those explanations make sense. I've done this kind of thing many times and the only time I ever worry about setting up a 3D environment is when the surfaces are in different planes. Every time I have done a job like this I go to the location and take a reference image from the projector's position. That is the only way I know of to get things properly lined up.

Not sure what you're getting at. si9nce AE doesn't have any genuine orthographic cameras, there will always be some perspective, regardless of how far you zoom out. Of course you can always emphasize it by animating the objects themselves or applying distortion effects to the 2D output. Generally, though, I would stay away from this kind of workflow in AE. It's simply not cut out for this kind of work. even the slightest bit of additional animation can make things look totally whacky at "infinite" zoom levels. This really sounds like you either need to think more about your project structure in AE or do it in a 3D program right away. More to the point it seems to me that at the very least you need to keep your proscenium setup in a separate comp and animate your objects in another with a more conventional camera, then mix everything as 2D layers.

Mylenium