787 out of spec or not?

[daddario]

Hi!
I'm having some confusion with the current source on my Fluke 787, which outputs a bit of garbage on the lower settings.
I've never hereinbefore been bothered by it as I really haven't used it with such small currents, but now I just wanted to finally clear things up, mostly for myself.
The error is pretty much the same for source and simulate outputs. Both the user-, and cal. manual state "Accuracy: 0.05 % of span".
Does that mean 0.05% accuracy of 20mA or 0.05% of the programmed value?
I did some measurements and this is the result:
The left number is the programmed value on the 787 and on the right, the value read from another DMM, rounded to a 100nA.
400uA    401,4uA
300uA   301,9uA
200uA   200,9uA
100uA   97,7uA
50uA   46,1uA
40uA   36,0uA
30uA   25,4uA
20uA   15,2uA
10uA   4,7uA

Above that it's pretty much spot-on.
If the accuracy is for the whole 20mA span, then a 10uA error is permissible at all times and everything is more than fine. But if not...