The very valid question whenever the topic of empirical standards comes up is "Why do you think you need it?" I am unconvinced that empirical standards are ever needed. Of course, I am also unconvinced that I am right in saying that! So who knows...? There is some evidence that empirical standards might be preferable in certain specific cases, such as the case in which a hydrogen atom is collinear or nearly collinear with an absorber and scatterer.
From a the persepective of "guy who writes programs" I suppose the fact that there is interest in having the capability of using empirical standards is itself good enough reason to include it.
One reason I can think of is when the unknown is in some sense close to a model, for instance, a glass compared with a corresponding crystal, a dilute solid solution compared with an appropriate compound, if one exists, or a nanoparticle compared with bulk. In that case, one might argue that Nature's calculation might be more accurate than FEFF, and that the use of a model will automatically correct for artifacts in the extraction of the (filtered) shell. I will grant that this is now a niche technique whereas, when I started <mumble> years ago, it was mainstream. mam