The models differ in their view of discourse semantic systems, such as conjunction, ideation, appraisal
In a) discourse semantic systems are not possible, as semantics is consumed with grammatical functions,
in b) they merely instantiate grammar systems in texts, and
in c) they are distinct systems, where grammar and discourse both contribute to meaning
 These 'discourse semantic' systems could easily be incorporated in Fawcett's Cardiff Grammar, as systems (functions) on the higher stratum of context, and as structures (forms) on the lower stratum. Given that some of them derive from Halliday's grammatics (textual cohesion), this may already be the case.
 This reflects Rose's ongoing confusion of instantiation with realisation (see elsewhere on this blog). Semantic systems do not instantiate grammar systems ('in texts' or anywhere else). Semantic systems are realised by lexicogrammatical systems.
Regarding the systems Rose mentions, while 'appraisal' is a genuine interpersonal semantic system, construed by the interpersonal grammar — the word 'discourse' is redundant — the others are a mixed bag. For example, Martin's logical 'discourse semantic' system of conjunction derives, at least in part, from Halliday's textual grammatical system of conjunction, and Martin's experiential 'discourse semantic' system of ideation derives both from Halliday's (non-structural) textual grammatical system of lexical cohesion and his (structural) experiential grammatical system of transitivity.
 As  indicates, integrating these diverse systems as Martin's 'discourse semantic' systems creates theoretical inconsistencies in terms of both metafunction and level of symbolic abstraction.
 This reflects Rose's ongoing confusion of semogenesis with stratification (see elsewhere on this blog). All strata make meaning, in the sense of semogenesis, but stratification is a means of parcelling out the complexity of language as a hierarchy of symbolic abstraction, in which meaning (semantics) is realised by wording (lexicogrammar). Does discourse "contribute" to meaning, or is it a semogenic process?