In recent years, the Mg ii h and k lines have become increasingly important for the diagnostics of magnetic fields in the solar chromosphere, motivated both by the advances in the modeling capabilities of their polarization and by the unprecedented spectropolarimetric observations provided by the CLASP missions. In this paper, we investigate the impact of the angular and frequency coupling in coherent scattering processes in the radiative transfer modeling of the Mg ii doublet polarization profiles. In particular, we aim to assess how the widely used angle-average approximation affects the predicted polarization. By solving the radiative transfer problem in 1D semiempirical atmospheric models, we find that this approximation is suitable in the absence of magnetic fields, but there can be measurable differences in magnetized models, mainly in the linear polarization profiles close to the core of the Mg ii k line. We then test the suitability of the approximation in more realistic scenarios, solving the radiative transfer problem pixel-by-pixel in a 3D atmospheric model resulting from a magnetohydrodynamic simulation and mimicking the degradation of the CLASP2 instrument, as well as in several 1D atmospheric models resulting from the Stokes inversion of the CLASP2 data. We find that the impact of the angle-average approximation is greatly diminished at the resolution of the CLASP2 observations. We also find that to suitably include angle-dependent effects in 1D radiative transfer modeling, it is sufficient to initialize the problem with the angle-average solution and then perform a few (angle-dependent) iterations, dramatically reducing the computational cost.