Look at the flow field, you may need to get images at several points in the cycle. They should show why you're getting fluctuations. If you're not sure post the images and I may be able to tell you where to look.
Enclosure dimensions are covered in literature, they're typically related to chord length and/or blockage factor. Look in both CFD and wind tunnel references.
A and B are both valid, but with some comments. For A I'd probably use 2d. For B a symmetry plane would reduce the cell count but still allow you to look at tip effects.
There is no "perfect shape". It's more a case of understanding why you're doing something. Blocks are easy, and work well for aligned aerofoils. The semi-elipse is good for varying angle of attack in one mesh: check the NACA tutorials/videos.