I think my recommendation to check for accuracy is to check where your tool will be. What I mean by that is that every added "layer" where there can be play is another place where you can get added slop. To check for accuracy of a DRO scale and to make sure it is reading properly, you should indicate the actual scale/reader head movement. However, that isnt always the exact travel of the tool. Once you add in the cross slide, the carriage, the tool post, etc (depending on what type of machine and accessories you have on it) then the 10 thou you moved is actually now 4 thou because of play, wear, etc.
As an example, I was troubleshooting a Bridgeport with no repeatability on the X. When I got there, the scale was repeating perfectly and reading accurately. I found the issue was when the table went to the extreme ends of travel and then back to zero, and the column/knee was worn enough that the entire knee/saddle/table was moving. Obviously the scale couldnt pick up that movement, but the movement was seen in the part.
Just some food for thought, since it was a puzzle that took me longer than I want to admit to piece together lol
Jon