I have been told that this cannot be done. So I will take some baby steps showing what is going on with this crazy plot of mine.
Neukum Et Al. have a Southern South America temperature reconstruction that I think is a pretty decent reconstruction to use. I am sure it is far from perfect, but it looks useful. The Satellite data is arguably the most accurate of averaging large area temperatures, though the data is not a surface measure, it should naturally average variations in surface temperature. The oceans, due to their larger thermal mass would also naturally average temperature fluctuations. So starting with the better data and working back in time makes perfectly good sense to me. We are after all comparing less than perfect indications of temperature.
There are a few other southern hemisphere and tropical temperature reconstruction that should be fairly closely related to ocean temperatures which should be a reasonable baseline for estimating past global temperature change not overly related to land use and ice sheet changes.