Quaternion to directional unit vector (LIGGGHTS)

Hi,
I am new to LIGGGHTS and have been having trouble finding information on how quaternions are calculated.
The particle template I have been using is multisphere, and I provide a template file that creates a pellet shape (picture provided). The trouble I've been having is when I try verifying using quaternions relative to computing orientation vectors that use individual particle positions. In the local coordinate system of each pellet, the pellet is orientated lengthwise along the z-axis. When my pellets are inserted with the template orientation ((0i + 0j +1j) = vec), both post-processes using quaternions (being calculated with quats as qvecq*) and particle positions agree. Yet, when inserted with a random orientation, they do not. This leads to my question: Are quaternions calculated with respect to an initial orientation? If so, is that orientation the template's orientation or the orientation of each pellet after it is inserted with a random orientation? It doesn't make sense to me for it to be the latter, as you would have to figure out how each pellet is orientated originally. Perhaps I'm not looking at this the right way, any advice would be appreciated.