Probably, because with modeling you can test a wide range of different hand+limb sizes and motions; you're not limited to the feedback from one tester, which, a

Probably, because with modeling you can test a wide range of different hand+limb sizes and motions; you're not limited to the feedback from one tester, which, as the OP notes in its discussion of the "German" grip, might have preconceived notions about swordplay not shared by the Greeks. Modeling can be informed by real-world use, though, certainly.