Rather than introducing bounding classes, it might also be useful to be able to poke cycle and boundary matrices directly into the spectral sequence on the first page, before any interaction with differentials occurs. These are also displayed to the user in read-only EXTMatrixEditor controls, and it would be simple to make them writable on the first page.
A good implementation of this would be a little delicate, though: how do we propagate this information to other terms? When building the KU HFPSS, for instance, I don't want to have to poke 2s into all of the boundary matrices of all of the terms containing eta-multiples.