Which to trust, home-built DROs or hash marks?

I recently completed fabricating and installing the Y axis of my home-built
DRO (using a 25 cycle per mm linear strip from Encodertech.com, which gives
a DRO resolution of 0.00039xxxxxx inches per jump of the encoder after
quadrature) using Art Eckstein's DRO program and interface plans (I used
linear strips instead of rotary disks because I dont have a lathe to turn
the bushings).
Now that I have the Y axis complete, there comes the question of
calibration. My mill is an Enco 16 3/8" Mill/Drill, and I found that when I
milled two flats onto a billet of aluminum such that the flats would be
1.500" apart (using the hash marks on the mill's Y handle to determine size)
using a 1/2" endmill that the following measurements resulted:
The Mill/Drill's hash marks read out 2.000" between centers (so the flats
should have been 1.500" apart)
My digital calipers (cheap chinese) said its 1.514" between the flats
My standard calipers (not dial) said its 1.515" between the flats
and my digital readouts (I calculated the incremental value using the 25
cycles per millimeter and calculated it out to be the 0.00039xxxxxx number
which I then plugged directly into the program, rather than using a scale
and needle for the calibration because I just ordered the scale on Thursday
and it hasnt shown up yet) said that the endmill moved 2.0164" (saying that
flats were 1.5164" apart).

Now the question comes up, what to trust? Should I just accept the 1% error
in the hashmarks on the mill and only use the DROs (the X and Z axis parts
should be arriving on Thursday)? Or should I investigate further to find out
if there's any other error in the system?

Thanks,
John McFadden