Your description of the algorithm I implemented is correct. I only did it for the [Mauna Loa CO2 data](ftp://ftp.cmdl.noaa.gov/ccg/co2/trends/co2_mm_mlo.txt), not temperature, since you seemed to be focused on the green curve. There are some missing values. It could be done with the data sets RealClimate used too. I ran their script and it produced the relevant output without errors, so I'm not sure what's going on with the one missing data set.