The classification of the less-divisible "free-form" signs (especially those seen in narrative and spatial arrangements) is definitely a challenge to linguists,

The classification of the less-divisible "free-form" signs (especially those seen in narrative and spatial arrangements) is definitely a challenge to linguists, but that doesn't mean that they don't consider them to be constructed out of semantic or phonetic units at all. For example, different objects and methods of description are frequently represented using the same classifier handshapes, despite how varied their placement and combination can be. On a similar note, the linguistic meaning of "word" is much, much fuzzier than native English speakers expect (letters surrounded by spaces). Trying to pin a particular level of structures across language categories becomes very difficult due to morphemic, phonetic, and syntactic processes that can break down word boundaries. (Even consider contractions in English for that matter. What makes don't one word when do not is two?) So the issue is, frankly, a mess, and many linguists tend to avoid using the term "word" cross-linguistically.