Two-scale models have been used for nearly two decades to estimate the scattering properties of rough surfaces described by their spectrum. Nevertheless, the solutions proposed in the literature present numerous differences. Through a rigorous treatment of the large-scale slopes we explain why some popular correcting factors actually should be omitted. Besides its simplicity, the resulting model is not restricted to situations where the large-scale process has small slopes. The proof is given for the general case of bistatic scattering from lossy dielectric surfaces.