The divider is quite unstable, and needs to be calibrated externally, quite often.
If you generate 0.1 and 1V on the 11V, you will have the non-linearity problem. The DAC is linear to 0.2 (measured) .. 0.5 (spec.) ppm, so 1V may be precise to about 2..5ppm, 0.1 V to about 20..50ppm..
Make sure to do a good zero on the Keithley (i.e. output 0V from the 5440A, and zero the 2015, then output 0.1 and 1V).
I thought, the 2015 would have 1.9x "overrange", for each range.. compared to 1.2x of HP DMMs.
Then you could test with 1.9V for the 1V range, 10V for the 10V range, both are same range on the 5440A, so only the linearity uncertainty applies.
Anyhow, the Keithley may be inside its 1yr. specification, but the 5440A reveals, that it has been drifting a lot since the last calibration.
A sign of that is that the 10V range drifted + 34ppm, compared to the others, which drifted in the negative direction.. Even if the 5440 would be outside its spec (maybe a few ppm), then all of these 10:1 ratios of the 2015 can be checked precisely.
The 5440 makes very accurate 10:1 ratio calibrations for those 4 ranges, I assume, even better than the 57x0As.
Frank