Best, HADCRUT4 and GISS all have the same limitations, the limited data they have to work with. The 1990 to present data is pretty good, but more by luck that skill. The average of large regions not well covered in a dynamic system will show its inadequacies. With both HADCRUT and GISS increasing their northern exposures 
they are diverging more from the satellite trends.
http://i122.photobucket.com/albums/o252/captdallas2/UAHpoleandtropicfrom2005.png
That appears to be pretty rare, the poles and the tropics all trending down, even for a short period. It may even be “unprecedented”
http://i122.photobucket.com/albums/o252/captdallas2/GISSpoleandtropicfrom2005.png
With GISS, one outta three is bad.