Yes, I think you are right lingle.. you can't learn about conversation without having meaningful participation in conversation. We set goals that are "behavioural" in the sense that they are observable and measurable, but we don't work on them outside of social context e.g. "what is eye contact?" worksheets. So the students have "targets" to think about, (e.g. one has a target to stop the interaction when they haven't understood to clarify, one has a target to use gesture and/or drawing if her meaning isn't being understood, another has a target to respond to a request for clarification by not only trying her message again but verifying that the other person now "gets it" etc).
The group have been meeting for a year, and initially, we had to give some basic structure - it was very much "tell everyone what you did at the weekend" or "plan a trip/discuss a trip", so relatively open ended. We sat them in a circle etc. Then we just let them at it, with a time frame e.g. 5 minutes. After 5 minutes (initially excruciatingly long), we'd have a chat with them about whether they'd enjoyed it and what help they felt they might need to enjoy it better. Generally, students were pretty good at saying what other people needed to do so they provide the feedback here. Then we might provide some support to help this happen, by joining in and modelling etc, then withdraw once they were going etc, then asked them was that any different.
At all times, if a student really wanted to say something that was not really translating to the group or we found hard to understand (either because the structure wasn 't there, or the form, or there was some social misunderstanding etc), we stepped in in a supportive role to help them get that message across: to "reveal" the message ^at the point that they had exhausted their own resources". Sometimes you need to establish what the person has in mind before you can work out what's going wrong.. then we would "rewind" and explain the breakdown and ask them to think of ways to solve it, modelling if they didn't have any (over the time the students have taken on themselves, we don't do this so much anymore, just a "what happened there?" prompt occasionally).
Literally, a year down the line, the students come in and have a free-flowing conversation about, well, anything. It is normal, unpredictable conversation. Earlier this week, a conversation that began about a family party that one member of the group had been to ended up being a conversation about religion that was jointly arrived at with all members participating in a completely logical, topic shifting way. The students do require support with some aspects, and I'm not going to pretend it's totally ideal.. it would be better to be able to replicate this with less familiar peers, then unfamiliar peers, and to do this in a variety of contexts.. but at least I feel they are having some opportunity in their day to just be in relationship with eachother and say all those things teenagers want to say without just failing and staying silent. We have seen some of these skills generalise to the mainstream, albeit in embryonic form e.g. one student now naturally looks at people and uses gesture because she has found it works, another is confidently stopping the teacher and peers in the whole class to ask for clarification on words.. there is a lot more to do... but I don't think that there's any way of learning how to do these things without, well, just getting stuck in and experiencing some actual success in a pretty average conversational context.
There is some info on supported conversation here: thomasland.metapress.com/content/cl3va94ade5ccvbe/fulltext.pdf - I think the text is freely accessible. This is very much about aphasia after stroke, but in terms of language disorder, I find a lot of it useful. I also find some of the work of Carole Pound on the politics of including people with communication disability meaningfully in society pretty interesting and a lot of it is relevant to people with moderate to severe receptive difficulties, especially as the prognosis for "fixing" some of these isn't amazing. This is not to say that I think understanding of language and the impairment itself shouldn't be worked on in its own right, particularly before the age of 20 when the brain is still maturing, but I do think it's important that self-advocacy and opportunities to have meaningful, supported interactions is embedded into therapy from early on.