What is the appropriate element for marking differing interpretations of grammatical analysis?

A project I'm working on includes ancient Aramaic texts where there can be more than one way to interpret the grammatical usage of a given word or phrase. For example, on a given line of a given manuscript, one authority says that the characters bm form a prepositional phrase consisting of two words:

<w type="prep" lemma="b">b</w><w type="noun" subtype="ms-cstr" lemma="m">m</w>

Another authority says those characters form a compound preposition and thus a single word:

<w type="prep" lemma="bm">bm</w>

In the digital edition, we'd like to retain and indicate both interpretations.

My first thought was <choice>. In its simplest definition it seems applicable: <choice> "groups a number of alternative encodings for the same point in a text." But the discussion of <choice> occurs within section 3.4 "Simple Editorial Changes" where it is shown as a way to group corrections (<sic> and <corr>) or regularizations (<orig> and <reg>) made by an editor. Here we're not dealing with an editorial change. Also, <choice> doesn't allow <w>.

My second thought is that the situation calls for some kind of critical apparatus. As I understand it, <app> is normally used to record textual variants from multiple witnesses. In our project, there is only one witness, but we have used <app> to record differing interpretations of what a given character actually is (authority A says it's an aleph, authority B says it's a yod). Our thinking is that this use of <app> doesn't stretch its intended semantics too far, because it still pertains to establishing the text. But grammatical interpretation is a different case. Instead of a textual ambiguity, we have a grammatical one. Also, <app> doesn't allow <w>.

There's <interp>, but that seems to be a kind of annotation. I don't view this situation as making an annotation, but rather as indicating two (equally defensible) ways of marking the same characters.

Is there an element I'm overlooking? Or should I circle back to <choice> and come up with a way to use <w> within it -- such as customizing my schema, or using choice/seg/w (which is valid but feels like a hack) instead of choice/w (which is invalid)?

Many thanks,

Gregory Murray
Head of Digital Initiatives
Princeton Theological Seminary Library