Rob,
In what context?
I mean, if they're talking about enterprise messaging, yes, that makes sense for a large organization.
But you have different contexts. A desktop or server may not always have connectivity to WebSphere, but would have access to MSMQ. It depends upon what you're trying to do. If you're attempting to buffer commands from a remote piece of equipment (like a handheld or something) before passing into the EMS, MSMQ is a viable and solid approach.
OTOH, if you're interoperating with a bunch of backend middleware built on MQ Series, that's another animal altogether.
My problem with an EMS is that if you don't it on every node, the network again becomes the weakest link and we all know that "The network is unreliable." It's the whole reason we have a persistent command pattern in the first place - because of that network thing that goes up and down.
If you haven't picked up EIP yet (The Hohpe book), definitely pick it up and give it a read through. Don't go in order, just pick a pattern and then walk through the related patterns and let it flow naturally around. The broker vs. bus chapter is good, along with a few of the earlier parts to just get an understanding of messaging concepts.
All of these things will help.
That being said, a WebSphere MQ transport shouldn't be too hard -- TIBCO was easy to build support for based on the Apache.NMS stuff used for ActiveMQ.
CP