About the NHC cone from the NHC website (the bold is mine):
"Historical data indicate that the entire 5-day path of the center of the tropical cyclone will remain within the cone about 60-70% of the time.
To form the cone, a set of imaginary circles are placed along the forecast track at the 12, 24, 36, 48, 72, 96, and 120 h positions, where the size of each circle is set so that it encloses 67% of the previous five years official forecast errors.
The cone is then formed by smoothly connecting the area swept out by the set of circles."
I've heard this before - that the cone is based entirely on historical forecast errors, and not on the model output for that particular storm. Why? It seems to me the cone would be more accurate if it was based on the unique synoptic pattern surrounding every storm, and its projected evolution.
For example, if the ECMWF, GFS, UKMET, and their ensembles were very tightly clustered through five days, then the cone should be very narrow, reflecting the high-confidence forecast.
On the other hand, if the ECMWF insisted on landfall in South Florida for run after run, while the GFS insisted on NYC and the UKMET was stuck on sending it into Belize after doing two loops, the cone would be very wide depicting the higher uncertainty in that particular forecast.