See On The Definition of Horizontal and Vertical Grids and Coordinates for Models-3:
The units for the coordinate system are meters, and it seems most exceedingly improbable that the SW corner of your grid is just over half a kilometer south of the map-projection center. This is most probably a “bad” file – where did it come from?
On the other hand, these two values are certainly within the bounds of reasonable round-off error of being the same thing – the newer version of CMAQ may well be using a bad round-off check. Why they didn’t use the I/O API’s carefully ropundoff-tuned coordinate-checking routine GRDCHK3 is beyond me.