Interesting, and glad you found a workaround. I suspect that something about calculating the mean/std err was not working with that missing data. I can't reproduce that here, so far, but I may not be missing enough of the same kind of data you were, or of course it may have been fixed since then.

--Molly