Which unit cell to use for dynamic structure factor in neutron scattering?

I’m trying to calculate the dynamic structure factor for inelastic neutron scattering, but I get a function which doesn’t obey certain symmetries of the crystal lattice and which changes depending on choice of unit cell.

For example, let’s say we have atoms A and B in a lattice like this:

ABABAB...
BABABA...
ABABAB...
BABABA...

We choose

AB
BA

for the unit cell, with A at (1/4,3/4) and (3/4,1/4), and B at (1/4,1/4) and (3/4,3/4), in lattice constant units. We can also choose

BA
AB

which should be valid since it’s just shifted by half a lattice constant. The dynamic structure factor (eqn 4.88 in Lovesey’s neutron scattering book) is:

$H^j_mathbf{q}(mathbf{kappa}) = sum_d bar{b}_d e^{-W_d(mathbf{kappa})+i mathbf{kappa} cdot mathbf{d}} (mathbf{kappa} cdot mathbf{sigma}^{j}_{d} (mathbf{q})) (M_d)^{-1/2}$

where $mathbf{kappa}$ is the momentum transfer, $mathbf{q}$ is the wavevector from the center of the Brillouin zone, $j$ enumerates the phonon branches, $d$ enumerates the atoms located at position $mathbf{d}$ in the unit cell, $bar{b}_d$ is the neutron scattering length, and $W(mathbf{kappa})$ is the Debye-Waller factor. The full one-phonon inelastic coherent cross-section (eqn 4.89 and 4.91) is

$frac{k^prime}{k} frac{(2 pi)^2}{v_0} sum_{mathbf{tau}} sum_{j mathbf{q}} delta (mathbf{kappa} + mathbf{q} – mathbf{tau}) | H^j_mathbf{q}(mathbf{kappa}) |^2 frac{1}{2 omega_i (mathbf{q})} [n_i(mathbf{q})delta(omega+omega_i (mathbf{q})) + (1+n_i(mathbf{q})) delta(omega-omega_i (mathbf{q}))]$

The problem is that, if I plot $| H^j_mathbf{q}(mathbf{kappa}) |$ for many $mathbf{q}$, I see that $| H^j_mathbf{q}(mathbf{kappa}) |$ is asymmetric about the (H,0,0)-axis, and this pattern flips about this axis if I switch from the first unit cell chosen above to the second one. It seems obvious to me that the dynamic structure factor should remain the same despite choice of unit cell. Can anyone let me know what I’m doing wrong or misunderstanding?