Hi Thunderbird12,
Actually, the left shift at 84 hours you speak of is still dramatically right of the 00z GFS run, per that same time interval. This is important because the two runs 12z and 06z show a demonstrative trend in the right direction (in general, forecasters like to see two consequtive runs to establish 'trend')

Note 00z GFS fix at 96 hours, centered on 00z Friday is approximately 89W by 17N.
Note 12z GFS fix at 84 hours, centered on 00z Friday is approximately 87W by 22N - which is entering the western side of the Yucatan Channel.

Using the 12z lat/lon differential to the 00z lat/lon for the same time interval, we can use...
c^2 = a^2+b^2
Solving for c, (60naut miles = 1 degree) gives an approximate distance of 325naut miles NE of the 00z run.

Edit -- note that this is not exactly correct. The Earth is spherical and the distance between points of longitude varies to a fair degree based upon latitude; thus, a great circle calculation is needed to find the true distance between two such points -- particularly over such a great discrepancy. For small distances, pythagorean principles work okay, but not as well here. The point is valid, just not the entire workings of getting there. --Clark

That is a huge NE (or right) shift in track... Now, granted, it is left of 06z as you intimate, but in terms of shear magnitude, the 12z left shift between 06z and 12z is far smaller than the demonstrative shift from 00z to 06z was to the NE..

For me, this is likely caused by many of the topics discussed to this point, regarding having an entity with a better physical presents in the model, so a better presentation out in time.

Now....none of this to this point is intended to presuppose any possibility that this could end up left - though personally, I don't believe that will be the case. Additionally, at hour 132, the 12z GFS is landfalling a powerful Wilma seemingly on an opposite telemetry that Rita took across S Florida; which is ironically even more right than the 06z for that same close approximated time interval! Weird ay?

btw: Any affects further up the East Coast is still open to debate in my mind because again....the Operational models seem to be having difficulty handling the amplification of the westerlies in the Ohio Valley by 120 hours. The last 24hours of runs had tended to back off on the amplification aloft, now in this 12z GFS run we've returned to a super deep negative anomaly cutting off and (potentially) veering the components into a deep southerly flow up the Coast... The 12Z GFS 'hints' that this 'could' capture Wilma and accelerator N, but doesn't quite pull that trigger does it... Hmmm... But, that is a really eerie solution because it is 'that' close to phasing Wilma into the overall evolution of this large negative anomaly...

Frankly, i'd love to see that happen wink wink

Actually, that was wrong...My apologies...disregard the paragraph headed with "btw:"...Apparently I stupidly forgot to refresh my images and therefore was unwittingly relying on yesterday 12z data for that paragraph.. But..seeing as I'm on the topic, still, not a certainty this would not affect areas farther N of the East Coast! ...time will tell.