Ref discussion at CA. There are various adjustments in HadCRUT and other global temperature series, and I removed those adjustments to obtain more accurate reconstruction (original obtained from Juckes et al supplementary):
raw
The original series is here, and this improved temperature series is here. Looks quite different, but no worries. Just imagine that the new one is the correct. Then take Juckes et al JBB standardized proxies, and check the correlations. They seem identical, 0.2810970282611 - 0.2810970282611 and so on. You'll get the same JBB INVR reconsruction using these temperatures.