If you accept that the Collaboration IS the meeting and the Interaction is what happened in the meeting, then the Interaction without the Collaboration is essentially meaningless. The Collaboration provides the structural context for the interaction.
I agree with Glassboy that the Aggregation is the wrong way round. In fact, I would question that the relationship between the "instance" of the meeting and its "type" (I'm using the terms a little loosely) is one of aggregation. We created a new relationship "instanceOf" to describe that link.
The relationship between the meeting "type" and its "instances" is analogous, in my view, to the relationship between types and instances for what we, here, call "Rosterable Item". A rosterable item is one where one can allocate a specific instance of a resource type to a specific timeslot. We plan, schedule and roster and then have actuals. We plan and schedule using resource types and roster using resource instances, actuals record the actual instances involved in the slot, the roster representing the forecast for the slot. I believe the analogy works well. You have resource type Monthly Meeting about X, which requires resource types of Chairperson, members of various types. You plan on "5" meetings, say. You schedule them for the next 5 months. Then you look for instances to roster against those types for those dates.
Thus, in order to model this correctly, you have to model the meeting type against the participant types and the business object types and then create relationships to the instances of all those types that represent the roster or actuals - depending on whether the date of the meeting is in the past or future.
While ArchiMate, as we said doesn't support the difference between types and instances, in section 8.2.2 Business roles, it does talk about the distinction between generic and specific business actors. In my view, the term generic is not correct, but the term specific surely refers to an instance. Elsewhere in the forum I have talked about the "generic" actually being a specification for a query that defines a placeholder into which a conforming specific instance can be substituted. So we would have "forum participant" (the placeholder) and "RWHurra" the specific instance.
HTH,
Paolo