No it's not essential. BSE gives you compositional info and u get more signal from heavier atoms. It's important to note that your material is NOT interacting with light and it's in fact interacting with electrons coming from your SEM filament. Below you can see where you are collecting BSE signal from (disregard the depth measurements in the figure since they are only associated with a certain accelerating voltage).
I agree with the above comment. However, if you surface is very rough, then from the hills of the rough surface more electrons will be scattered, thus collected (In SE image you will have brigher hills, darker creeks). So, rough surface will effect your secondary electron image but should not effect BSE image.
I think it is not correct to show a sketch of a flat surface in order to discuss the missing impact of a rough surface. It is quite easy to imaging what happens if a beam hits a microscopically slanted surface segment. The backscatter behavior is definitely changing, see e.g. Reimer (1987). You can even combine the images of e.g. four different BSE detectors (quad) to determine the surface topography, see the attached images. There simply a part of the EBSD (raw) signal has been used as BSE detector. Here the entire sample is tilted which certainly falsify the signal, but this technique is definitely used for 3D imaging of (horizontal) sample surfaces. Look for Shape of shading by BSE (e.g. Beil and Carlsen; Mashine Vision and Applications 4 (1991) ,271)
The first image is an BSE image, the second the derived 3D surface profile.
It depends on your objective with the BSE analyses. Smooth surfaces can be useful to avoid shadow zones in your surfasse that becomes difucult the ariving of the BSE electrons to the detectors.
In my case, sometimes I use BSE in rougous surfaces without problems.