The graph above is the USHCN monthly data – final minus raw for all stations. The graph shows the usual cheating, adding 1.8 degrees on to 2011 relative to 1900.
Now look at 2012 – the relative adjustment is much smaller.
It makes no sense algorithmically to have a discontinuity like that between 2011 and 2012. The only thing I can think of is that the hot weather in July allowed them to cheat less than they normally do in order to declare a record maximum.