I'm fairly certain that there exists exactly one solution, and it is a technicality that goes against the spirit of the embedding property.
Currently, verbs are greedy, in the sense that if something can fill them, it will fill them. Because the embedding property says nothing about where or how the clause generated from using the embedded property is used, we can say that underfilled ꝡä's are only allowed in the final slot of a predicate.
A clause that has *any* underfilled subclause would count as underfilled as well, consider this example, where the generated clause is also only usable in the final slot of a verb:
V3 N ꝡä V2 N N -> ꝡä V3 N ꝡä V2 N N
Obviously this solution goes against the spirit of what the embedding property is *actually* trying to upkeep.
We can't make subclauses non-greedy, so that underfilling is always the default, because they would stop having the same meaning once they are put in any non-last slot. The previous solution is still a requirement for this to function.
There exist no other solutions because verbs are going to be either greedy or non-greedy. If verbs are greedy, subclause underfilling is impossible. If verbs are non-greedy, subclause regular-filling is impossible. In both cases, we need something that is capable of indicating where the current clause is terminated.