I generated a list of all stations with no fabricated data from 1990 to 2013, and generated raw and final average temperatures for those stations. I also calculated the average of all fabricated temperatures at all stations from 1990 to 2013, which comprise more than 40% of the recent USHCN data set.
As you can see below, the slope of the complete final (orange) is slightly larger than the complete raw (blue), due to TOBS adjustments, but the fabricated data is completely out of line after 1998, about 0.8 degrees warmer.
You may recall that I have previously observed a discontinuity after 1998. This is a big part of it.