COOR ANAL can handle PBC with orthogonal lattice types, but only computes MSD for single atoms, not the molecular COM.
By "small", I meant less than ten or so lipids; your system seems okay.
Note that since there is both a short time and long time diffusion, one must care in the fitting to discard either the short time or long time data. Plots of r^2 vs t, esp. averaged over molecules, should make it clear where the slope change occurs.
