Multiscale composite models based on the Bragg theory are widely used to study the normalized radar cross section (NRCS) over the sea surface. However, these models are not able to correctly reproduce the NRCS in all configurations. In particular, even if they may provide consistent results for vertical transmit and receive (VV) polarization, they fail in horizontal transmit and receive (HH) polarization. In addition, there are still important discrepancies between model and observations of the radar modulation transfer function (MTF), which relates the modulations of the NRCS to the long waves. In this context, we have developed a physical model that takes into account not only the Bragg mechanism but also the non-Bragg scattering associated with radio wave scattering from breaking waves. The same model was built to explain both the background NRCS and its modulation by long surface wave (wave radar MTF problem). In part 1, the background NRCS model was presented and assessed through comparisons with observations. In this part 2, we extend the model to include a third underlying scale associated with longer waves (wavelength ~10--300 m) to explain the modulation of the NRCS. Two contributions are distinguished in the model, corresponding to the so-called tilt and hydrodynamic MTF. Results are compared to observations (already published in the literature or derived from the FETCH experiment). As found, taking into account modulation of wave breaking (responsible for the non-Bragg mechanism) helps to bring the model predictions in closer agreement with observations. In particular, the large MTF amplitudes for HH polarization (much larger than for VV polarization) and MTF phases are better interpreted using the present model.