Puerto Rico does offer up some decent mountain ranges. I remember years back when I chased Hurricane Hugo. First we were on the north coast, then traveled down to Ponce, then crossed the island northward via a nice highway system in place there, and finally back to Luquillo/Farjardo area. I cannot however recall a decent sized tropical cyclone with a good size wind field actually torn apart from the island. I good hurting to a storms symetry - certainly a disruptive force. I have found that a storm will tend to "jump" or relocate more easily when passing near or over P.R., than what I call "the rock" ( Hispanola ).
Tonights GFS run from 0Z certainly does join the camp of previous Euro, UK, NOGAPS, and GFDL models. Also contrary to recent GFS runs, this one not only stalls the cyclone, but really intensifies it as well. In noticing the NAM, I realize it too has been very consistant run to run, however seems to fast. I think the GFS solution is certainly plausable due to the COL created by a lifting trough and the building ridge. Timing would dictate if such a stall were to occur east of Andros Island, farther North or South along the Florida coastline, or even in the S. Bahamas. I just do not see recurvature as a likely event with the westerlies retreating northward while simply leaving a week hanging ocluded surface trough. More likely to me that recurvature would occur, would be for 92L to be severly disrupted by Hispanola.