ad 3.: combining several independent structures in a single model is not a good idea. The solution methods are not being developed for this, at least in ATENA (and probably in most civil engineering FEM packages). Even if everything is linear (and the geometry is NOT in ATENA), stiffness differences make the combined system worse conditioned compared to the separate ones. Moreover, it can be hard to identify which of the separate structures has a problem.
Ad 4., 5.: Looking at the model with 1 and 3 surfaces, I see no connection among the 3 surfaces (just between the 3 and the large one)?
I am asking our colleague responsible for the ATENA-GiD scripts if/how one can connect the rotational DoFs between neighbouring 2D shells.