Sorry greg; it took me a lot of time to complete my reply to Stephen's post so I couldn't check that your response had arrived. I guess I have already answered your question (Can you explain a bit more about the overall purpose of your solution. What are you trying to do?)
As I specified in my reply, the main purpose of this type of design is to have less setup overhead. If I need to develop custom mapping, that would require development for each new file type / format. I need to deliver a solution where only a handful of mapping guys could deploy schemas & maps (& pipelines) to BizTalk Server 2006 machine, add some new map definitions in my system and the system just does its job (i.e. executes the maps for the files they are configured for).
I have also considered multiple Receive Location setup. But with the number of maps this solution needs to support (2000+ maps), it seems an overkill (I may be wrong here). Setting up receive location with each new map defined in my system is also an overhead (though a one-time effort to write some code to setup necessary ports, locations & pipelines)