Structured Body: help

Dear all,

I'm vary happy that Named Graphs has been included in the specification as
I think they are very useful.
I'm in the process of modeling the Pundit (http://thepund.it) triplestore
to conform to the relatively new specs.

I have a couple of questions/doubts:

"It is RECOMMENDED to have a dereferencable resource that contains the
serialization of the graph, rather than to use Named Graphs."
Does it means that Named Graphs, when used as body of an annotations should
have URL and that this URL should be dereferenced to the triples contained
in the named graph? If it is so I think it is a very good approach.

Regarding the Structured Body (
http://www.openannotation.org/spec/extension/#StructuredBody): it seems a
very strange approach to me.
If there has already been a discussion about it please point me to it...I
cannot find anything in the ML archive.
I'm referring to the example that uses X, Y, Z, etc. The question is: how
do I distinguish triples that belongs to the structured body from other
triples that state something about Y, Z or X?
May be the answer is in this sentence "Implementations that expect a
resource with a serialization separate from the Annotation graph may not be
able to store structured resources"? Does it mean that resource X, Y and Z
should not be involved in triples other than the ones included in the
structured body? Does it also mean that two different annotations cannot
have structured bodies that talks about the same resource (e.g. Y)? Isn't
it a bit too restrictive :-) ?

best,

Christian