Brett -- something like the MM5 is a mesoscale model, whichever flavor of the MM5 you prefer. The globals are the GFS, ECMWF, NOGAPS, UKMET, and so on. The statistical models take their output and modify it, whether as a simple mean or by applying some sort of regression/minimization technique to it, and include the FSU Superensemble and the various ensemble means.

I'm just primarily concerned about the ability of the global models -- which don't have quite as good resolution as the mesoscale models -- to pick up on the fine-scale features that will determine/influence Dennis' track. They may well prove to be perfect with the track, but as with any model output -- including the mesoscale and statistical model -- you've got to remain skeptical and use it as it is intended to be used: as guidance.