In microstructure models using CA, I have not yet found a deterministic relation for the grain orientation. It seems they just use random orientation for each grain.
During my Ph.D., non-random orientations were tested. I cannot recall right now how well are those tests documented in my thesis (online on RG) but I am sure that I did it.
Actually, right now I do recall that the neckling process was tested and modeled (it should be there). In necking the next generation of nuclei was taken with a difference about 10 deg of the orientation difference between mother grain a new nucleus (hence maintaining the high angle minimum between grains)
You can no lay around with mentioned processes with the open-source software provided on RG -- it is rewritten in C++ and Qt from the original code in a specialized language.
Orientation as a characteristic of a grain is the input and thus depends on how you set it. You might prescribe the absence of texture to the simulated microstructure (random orientations) or say that each grain of the polycrystal has the same orientation (it won't be a polycrystal then, however). Hope it helps.