It appears that there are a small number of stations in the file that have normals but no standard deviations. Looking at just those whose 'Normals source' is 'Data' it appears there are six where it should be possible to calculate the standard deviations (since the normals are being calculated from the 1961 to 1990 data): 104880, 606700, 680140, 829830, 835790, and 961090. In each case it would appear that there's enough data in the 1961 to 1990 period to meet the '15 years of data' criteria in Brohan et al. 2006. And given that there are normals generated from 'Data', it would appear that you could get the standard deviations.
Is there any reason why I can't calculate them and safely include these six files? Both your code and my code currently discard these stations because the standard deviations are missing.
They've been kind enough to reply:
When the station standard deviations were calculated, a slightly different criterion was used to judge what was a sufficient number of years: 16 years was used instead of 15 years. There is little harm in relaxing that criterion as it appears only to affect these stations.
I've followed up asking if there was a rationale for 15 for normals, but 16 for standard deviations. But looks like those stations can be safely included.
Update The Met Office has replied to the 16 vs. 15 question.