Hi
I don't think this principle is the general fitting answer and it always depends on the model intention, the camera and lighting options etc. etc.
In principle it is often most intelligent and sufficient to work with a simplified model version of an environment. - by means of suitable real photo texturing material, this can often "fake" the impression of a real situation better than even the most elaborate modelling work - at least as long as it's just the scene background 😉

Exactly like a native SkUp model is 3P perspective, your real photo will always be a suitable 3P perspective! When photographing you can already consider which edges might be useful for you later in SkUp - Useful for photomatching are pictures only then when you can adjust the red and green axes later on.

And only for graphic reasons for example and only if needed:
2P perspective instead is always an "artificial" alienation of the image output above only - this output option is still available in SkUp as well as in any renderer or also later via post pro in any image editor ...