Jamestown deaths and drought: the archived numbers give r = -0.11, not the published -0.71

Stahle et al. 1998 report a correlation of -0.71 between Jamestown percent mortality (Earle 1979; 1608 to 1624 plus Roanoke 1586) and reconstructed July PHDI. Using the 18 values in their own archived file we get r = -0.11; with the atlas's summer rain the same year +0.21, the year before -0.43 (p 0.04, one of four tries). Many 'annual' mortality values are averages over several years (the data note warns of this), so the published figure may come from grouped periods. Worth asking rather than assuming an error.