You're doing the mapping part, but you're
		leaving the length of the valence lists underdetermined.
		For example, your nouns have SPR lists that have at
		least one element, but nothing says they can't have more
		(which is why 'la la la hundo' etc are parsing as NPs).
		You've also neglected to say that the transitive verbs
		have exactly one COMPS element, and to say anything about
		the SPR value of verbs.  (Doing so would make the 
		bare-np-phrase work without constraining it to be [HEAD noun],
		and also rule out 'la mortis' as an instance of head-spec.
		You note that 'la' etc can be used as standalone NPs
		in certain circumstances, but in order to get the semantics
		right in such cases, we'll surely need a different non-
		branching construction.

		Pred names for verbs should be in the citation form,
		not the past tense.

	
	The main distinction between determiners and other nominal
	dependents is whether or not they can iterate.  You're right
	that we'll need more than just def_q_rel etc to get the semantics
	of possessives


Response to concluding comment: it sounds like you didn't ask
questions soon enough :)

Interesting observation about the labels and a free word order
language.  The labels are just for convenience, but I can see that
we might need to think a bit more about their order in the file
to make them most intuitive for free word order languages.