One thing I will say: IFG 3rd ed. (2004) was a result of extensive addition by Christian of Halliday's IFG 2. One thing I have noticed is that Christian tends more to notional coding than Halliday, swayed more by semantic similarity, while Halliday is more grammatical in his coding.
For instance, there is a new example in IFG3 involving "talk .. .about" being a verbal process, while the previous page (and IFG 1) list"talk" as behavioural.
In any case. the BNC lists one example of "mourn" with a projection:"The steward was not left to mourn that his bottles found no custom"Admittedly, one instance is not substantial. But this sentence reads ok to me at least.The principle as I understand it is that the verb has to have the POTENTIAL to project verbally to be considered mental or verbal. Otherwise it may be behavioural (as in "talk").
 Matthiessen, like Halliday, advocates taking a trinocular perspective on the grammar, looking at it 'from above', 'from roundabout' and 'from below'. This is in contradistinction to "notional coding".
 The verb talk can serve as verbal process or behavioural process, depending on how it functions in a clause. O'Donnell again makes the error of taking the verb itself as the point of departure for determining the function of an element of clause structure.
 This example is a single clause with an embedded fact — (the fact) that his bottles found no custom — not a clause nexus of projection.
 The principle that O'Donnell "understands" is entirely of his own making, and one source of his confusion. There are subtypes of both mental and verbal processes that either don't have the potential to project, as in mental processes of perception and emotion, or are relatively unlikely to project, as in 'targeting' verbal processes.