As I showed in my Hub, when you consider that metric, the agreement of models and observations is quite reasonable. (hubpages.com)
As I understand, climate sensitivity can be obtained both from models and observations. (skepticalscience.com)
Including the effects of natural external forcing factors has a relatively small impact on our 1950 — 2005 results, but improves the correspondence between model and observations over 1900 — 2005. (link.springer.com)