In general, yes, the Scd for each of the 2 C-H vectors on a given C atom are averaged, as they cannot be distinguished experimentally in most cases. The exception is C2 of each chain (C22 and C32), especially the beta chain (C22); the C-H vectors are inequivalent and can be distinguished by the NMR quadrapole splitting experiment. Computing the Scd for the 2 vectors separately is a test of convergence; if <Scd> (average over time) for 2 C-H vectors at a given chain position don't agree that well, the simulation may need to be run longer.
