David Rose posted (apparently unsuccessfully) to sys-func at 11:30 on 18/2/14 and then (successfully) to sysfling at 15:41 on 19/2/14:
My point is that this is a description of the meanings made in the grammar. A point you are making (I think) is that relations between meanings in grammar and meanings in discourse are not one-to-one, but many-to-many. You see the semantic stratum as organising grammatical choices, to realise higher level meanings. This is quite different from the model I have been questioning, in which semantics is held to be a mirror image of the grammar. But it is not so far from the theory of discourse semantics.
 Ideationally, the lexicogrammar construes experience as meaning, where 'construe' means “construct semiotically, transform into meaning” (Halliday 2008: 2); and in terms of stratification, the lexicogrammar realises meaning.
 In the SFL model that challenges Rose, semantics isn't "held to be a mirror image of the grammar". In congruent realisations, sequences are realised by clause complexes, figures by clauses, and elements by groups/phrases, for example. But in the case of grammatical metaphor, the realisation relations can vary considerably, as when a semantic figure is realised, not as a ranking clause, but as participant in the function structure of a clause, or as the Qualifier of a nominal group that realises a participant; and so on. Some helpful quotes from the ideational perspective:
Halliday & Matthiessen (1999: 26):
Thus when we move from the lexicogrammar into the semantics, as we are doing here, we are not simply relabelling everything in a new terminological guise. We shall stress the fundamental relationship between (say) clause complex in the grammar and sequence in the semantics, precisely because the two originate as one: a theory of the logical relationships between processes.Halliday & Matthiessen (1999: 237):
Of course, what we are recognising here as two distinct constructions, the semantic and the grammatical, never had or could have had any existence the one prior to the other; they are our analytic representation of the overall semioticising of experience — how experience is construed into meaning. If the congruent form had been the only form of construal, we would probably not have needed to think of semantics and grammar as two separate strata: they would be merely two facets of the content plane, interpreted on the one hand as function and on the other as form.