The GFS outputs for the entire globe. Why does NCEP care that much about a margin of error of 50 miles on a storm in such a small geographic area? Not making excuses here, it's just that New England snowfall accuracy in the coastal plain is just one of millions of chaotic events going on at any given time that the models need to predict.
That being said, if the US put the same resources into weather modeling like the ECMWF does, we would be better for it. We used to be the best at this, and we no longer are.