The effect of angular structure differences between measured and best-fit analytical phase functions of the equivalent backscattering ratio on calculated reflectance values was studied and shown to be significant. We used a Monte Carlo radiative transfer code to check the effect of choosing different analytical (several Fournier- Forand (1994) and Henyey-Greenstein (1941)) phase functions with backscattering ratios identical to the ‘classical’ average Petzold function. We show that the additional variability of the resulting water leaving radiance is about 7% (4% between the Fournier-Forand functions themselves) for most scenarios. We also show a previously unknown maximum of the discrepancy (up to 10%) for highly scattering waters. We discuss the importance of relative differences in phase function for different angular ranges to this maximum and to the behaviour of the discrepancy as a function of solar zenith angle.