In my opinion, the reference density should be the one of the unperturbed flow, so it should be the same for all the simulations if temperature is not varied.nFurthermore, for supersonic flow (M < 5) pressure-based gives a right performance, so do not worry if you used that solver. n