Looking at the models and weather boards, my strong suspicion is that the NWS is predicting much less snowfall than we're seeing in many models due to ratios and sleet, mostly, plus some melting. It looks like it has to do with the models reporting "snow that falls" from the sky (without melting and accumulating at a 10":1 ratio) vs. looking at how much snow will actually accumulate on the ground, given poor snow/liquid ratios in what is falling, combined with possibly significant melting/compaction on the ground and assuming some of the frozen precip is sleet, which has a very high density (low ratio of frozen depth to frozen mass).
Looking at one model kind of shows this. Below is the snowfall graphic for the high resolution 3 km NAM model assuming 10:1 ratio, showing 6" or more of snow for I-95 and NW of there, whereas the next graphic shows the Ferrier algorithm for snow, which takes into account low ratios and sleet, but apparently not melting, which may be why the NWS numbers are even lower than the Ferrier ones in some locations. Huge difference in the snowfall maps from these two different algorithms and I'm pretty sure this is the main reason for the discrepancies between the models and the forecasts (and the low ratios/sleet/melting are very connected to the warm air aloft emanating from the warm ocean).
Keep in mind, however, that if we see 1" of sleet, that's the equivalent of about 3-4" of snow, so even if it doesn't look as pretty, it contains the same mass of frozen precip, so it'll be just as hard to drive in, plow and shovel as 3-4" of "standard" 10:1 snow/liquid ratio snow (except for the visibility, which will be fine in sleet). However, it's also just possible that the NWS and local forecasters will be wrong and the models will be more correct.