Myself, I would have treated the two 'which' clauses as nondefining relative clauses, and thus as part of the nominal group structure, and not as part of the clause complex structure.
and then again on 15 September 2013:
And yes, I did mention that I disagreed with the treatment in the example of nondefining relatives being related as part of a clause complex.
My reasoning is that they only ever occur immediately after the noun (with possible interceding Qualifier) with which the relative pronoun is grounded.
And semantically: They do not extend on a clause, but on a participant.
I don note the possible counter example of postponed nominal relatives,
But note that this case is only possible where the postponed relative is in a defining relationship. The above sentence corresponds to:
- The man is here who you were waiting for.
- The man who you were waiting for is here.
- The man, who you were waiting for, is here.
 O'Donnell's stated reason for treating non-defining relative clauses as 'part of nominal group structure' is 'that they only ever occur immediately after a noun (with a possible interceding Qualifier) with which the relative pronoun is grounded'.
The domain that non-defining relative clauses elaborate is not restricted to the nominal group; they can elaborate the whole of the primary clause or some part of it that is more than the nominal group' (Halliday & Matthiessen 2004: 400). For example:
He often accuses others of his own shortcomings, which is hypocritical.Here the non-defining relative clause which is hypocritical elaborates the whole primary clause he often accuses others of his own shortcomings. Clearly, it is not a constituent of the preceding nominal group, as any attempt to analyse it as such will demonstrate. Therefore, O'Donnell's position is based on a false premise.
For some reasoning about the difference between non-defining and defining clauses, see:
 Relative clauses generally elaborate the primary clause rather than extend it, though there is one group that strictly belong with extension. See Halliday & Matthiessen (2004: 402-3) for the reasoning.