The 00Z runs of the NWS and ECMWF models have essentially the same solution; Irma coming up the spine of FL, albeit the ECMWF code is about 12 hours faster than the NWS code. The official forecast from the NHC at 126 hours is very close to the NWS model with the position just south of the FL keys (2AM Sunday). One saving grace is the typical track errors at that forecast length are 200 miles. Staying out over water in the Atlantic, however, is appearing to be less likely. The forecasts from Wednesday onward should start to become much more reliable. Essentially, the NHC only puts out a 5 day track forecast (with an envelope) because there is little or no accuracy beyond that point. The global numerical model is run out to 15 days 4 times per day - often there is not a lot of skill beyond about 5 or 6 days, depending on the situation. Ensembles, based on slightly altered initial states, are run to provide some feel for when accuracy starts to diminish.
Here is the latest track, with an envelope, from the NHC.
HURRICANE IRMA