I believe in in none of those cases does nei or no'a "work". I don't consider that sei-se'u constructions can "see" the sentence they are talking about it. It would make about as much sense as:
lo mamta cusku lu vo'a nitcu lo nu klama lo zarci li'u
and claiming that "vo'a" refers to "lo mamta" (contrariwise, "lo mamta cusku lo se du'u vo'a nitcu lo nu klama lo zarci" does work)
--gejyspa
CLL 19:12 contradicts you on this point:
Since a discursive utterance is working at a “higher” level of abstraction than a non-discursive utterance, a non-discursive utterance cannot refer to a discursive utterance. Specifically, the various back-counting, reciprocal, and reflexive constructs in selma'o KOhA ignore the utterances at “higher” metalinguistic levels in determining their referent. It is possible, and sometimes necessary, to refer to lower metalinguistic levels. For example, the English “he said” in a conversation is metalinguistic. For this purpose, quotations are considered to be at a lower metalinguistic level than the surrounding context (a quoted text cannot refer to the statements of the one who quotes it), whereas parenthetical remarks are considered to be at a higher level than the context.
Thus, it is possible from within a "higher" context to refer to the contents of a "lower" context. This is why it is possible to refer to the contents of a lu-quote, but not for its contents to refer to the bridi in which that lu-quote is a sumti. Likewise, a sei-clause can use ri and other anaphoric devices to refer to what is outside it, because what is outside it is on a "lower" level.
Metalinguistic "height" according to CLL:
SEI..SEhU, TO..TOI > ordinary bridi > lu-quotes