0.1% cumulative?
Well that is bad for resistor testing though it is acceptable for many other uses.
Edit -
but the ADC is a stair step, not a slope. The slope runs up the upper corners. Everything between the slope and the step below is lost to error. That's how the data gets captured.
For resistor testing wouldn't it be more correct to add 1/2 step to the output to at least put the stairs through the slope and make the error +/- instead of --?