Okay, well I was not able to figure it out. This post seems relevant, but only to the point that the mass zones (off of which I believe the analytical spaces are created) is an automatic process once there are masses and mass floors in place.
The gbXML comes out error free (as reported by an online validator), but as you can tell when looking at the model, lots of gbXML seem to come out weird. E.g. the ceilings come out as roofs, interior walls come out as exterior walls, etc. Fairly frustrating experience trying to use conceptual models for building the gbXML model, although it does feel like the best route to work around Revit's shortcomings in developing functional gbXML models. The tutorials I've seen online typically deal with masses that represent entire buildings. The one conceptual mass does not include interior partitions, but will output a functional gbXML.
I think the problem with the approach I am taking is that each mass is treated like a building (maybe), so each mass includes its own set of exterior walls, roof and slab. Not sure if there is a simple fix to this, or if I am approaching this the wrong way.
Edit: Upon further investigation, I do not believe I can even change the conceptual mass wall type from anything other than exterior. I believe this is the data that will drive the wall assignment in the gbXML file, so unless I can change the mass zone wall type, I am stuck. I think that maybe creating a mass that represents the shell of the building and nothing else is the best I can do. However, I don't know if the script you wrote to combine the gbXML's will function will this kind of output. I'm glad I'm at least documenting my workflow somewhere, so someone else doesn't stumble through the same mistakes I did.
