It seems odd that the values are so high in the recent period of this series and could conceivably be instrumental data , but would have to check. The scaling of the data we used to produce the Crowley curve that formed one of the lines in our spaghetti diagram (that we put on the web site under my name and made available to NGDC), was based on taking the unscaled composite he sent and re-calibrating against April - Sept. average for land North of 20 degrees Lat., and repeating his somewhat bazaar calibration procedure (which deliberately omitted the data between 1900-1920 that did not fit with the instrumental data (remember his data are also decadal smoothed values). In fact , as we were using summer data we calibrated over 1881-1900 (avoiding the high early decades that I still believe are biased in summer) and 1920 - 1960 , whereas he used 1856-1880 and 1920-1965. Of the precise details might differ - but the crux of the matter is that I suspect one of the Figures you show may have instrumental data in the recent period - but not ours. If you say exactly where these series came from I can ask Tim (who will have done the calibrations) to check.
Madness at #AGU2018 – probability based tornado warnings
33 minutes ago