Your point about close up objects concerns me as well. My tests with a standard head showed this up but I was hoping that a dedicated pano head would go some way to sorting this out. Am I being a bit optimistic on this? Is there still paralax issues even with a pano head?
This is the question of accuracy: how accurately you measure the location of entrance pupil and how accurately you can adjust the camera position.
It is not easy to achieve high accuracy.
1. The location of the entrance pupil depends on the focusing distance as well, with many lenses. For example the Canon 50mm f/1.4 changes the location by focusing; not much, but enough to cause small errors. This fact makes fixed adjustments useless. RRS offers (or offered) some stuff dedicated to certain lenses; this is dubious, to say the least.
2. In order to verify the adjustment in the initial measurement, one has to shoot some close AND far objects in the same shot (and two shots are required for one comparison). However, where should be the focus? In order to judge the adjustment with single pixel accuracy, the focus has to be on the subject - but on which one? One of them will be OOF to some degree.
3. Zooming lenses change the entrance pupil with zooming.
4. A tiny error in mounting the camera, i.e. if the axis of the lens is not parallel to the rail when measuring or using the lens can cause much bigger error than a small error in the adjustment itself.
High accuracy in the adjustment and high pixel density for downresing (and hiding tiny errors) is required for good indoor architectural panos. I often laugh when seeing the proudly presented samples on forums - in a fraction of the real size.
Btw, if it is necessary to change the focusing between the frames, then only a decent stitcher can make a good pano, for changing the focusing means changing the focal length, i.e. changing the angle of view, which is a critical parameter of the stitching.