The trimming of the 6 MSB of the DAC is very coarse, about 2500/5000ppm for 2R or R, respectively.
Therefore, trim resolution and minimum linearity is worse than 10ppm, at best.
So the root cause could simply be a mis - adjustment.
Otherwise, I would first check, if the switches inside the potted case provide exact 0V and 5V to the six 2Rs.
Simply measure on pins 2,3,4 and 13,14,15 of the connector, when switching different values.
A nicer and more precise R2R - DAC is inside the Fluke 3330B, maybe interesting for understanding trimming scheme, and so on.
Frank