I also found this explanation:The BAM-series models take the GFS analysis & forecasts and computer trajectories from the storm's initial position out to 120hr. The BAMS uses the 850-700mb layer, the BAMM 850-400mb, and BAMD the 850-200mb layer, with a small correction for the Beta effect (due to Coriolis; generally <2kt in forward motion towards the northwest). Note that the GFS has been consistent in decoupling the low-level and upper-level circulations of Franklin since its inception, taking the surface features to the north and east while meandering the upper-level features back to the west and southwest.

The two BAM-series models we've been seeing are the BAMM and BAMD, neither of which include the lowest levels to any large degree in their forecasts. Thus, their forecasts are largely going to be a representation of what the GFS views as the mid & upper-level steering pattern. Given what I noted above, it's not too much of a surprise to see them steer the storm more towards the west and towards land. This suggests that the GFS has either too strong a ridge, too weak a shortwave trough, or some combination thereof. No matter the forecast, the GFS solution is a rather dubious one for Franklin (unless you believe the storm will become sheared apart and fizzle out through time, which wouldn't be a bad thing necessarily), making the BAMD and BAMM solutions rather dubious at best themselves.

You can't entirely discount all of the forecasts, but I'd rather go with a model that represents the storm better -- such as the NOGAPS or even our MM5 (which uses the GFS initial conditions to start the model, but diverges completely from there) -- when making a forecast. Both of those models -- and most of the others -- agree on taking it north and then northeast, with varying degrees of speed through time.