Here is the current state of my model, it has a feedback with the albedo, and inertia. The sum error^2 for 1000-1900 is 4.49.
What I don't have is volcano data from 1998 on.
What I had to guess on was the increase in effect of CO2 from 1998 to 2012.
Both of those is why there is a sudden jump in calculated temperature at the end that exceeds bot the slope and the actual measured temperature.
Things I have learned, though I have suspected prior is that the solar variation has to be greater than what we have observed, and that there may very well likely be an error in the measurements of the TSI data, and reconstructions. It is possible that there is more atmospheric sensitivity to solar variations, but my best guess is that the solar variation is varying more than we think.
Also it seems that the dips in temperature due to volcanic activity don't always have a lag in the return response.
There are also some other factors that seem to cause a change, or lack of change at different time periods that I am not accounting for. It could be error in one or more of the data due to proxies, or just error in my model that I have to work out.
Orange line is the recreation, blue line measured.