Thanks for your post.
The calculation of AOD in AOD_DEFN.F includes all aerosol species and does not disregard BC and dust.
When computing AOD using the IMPROVE reconstruction method, the difference between the species you listed (SO4, NO3, NH4, OM, and sea salt) and all other aerosol species is that the extinction computed from the listed species includes a humidity correction while the extinction from all other species does not include a humidity correction. The computation for ammonium sulfate is performed in lines 411 - 427, the computation for ammonium nitrate is performed in lines 429 - 445, the computation for OM is performed in lines 447 - 461, the computation for sea salt is performed in lines 464 - 471, and the computation for all other species without a humidity correction is performed in lines 473 - 477. The visibility indexes used for all calculations can be found in AERO_DATA.F
The images you posted likely point to an underestimation of windblown dust in your simulations, although comparing the AOD approximation using the IMPROVE reconstruction method to the MODIS AOD product also has some inherent uncertainty.
Note that the use of the IMPROVE AOD estimation method was discontinued in CMAQv5.3 and aerosol extinction information including AOD can now be obtained from the PHOTDIAG files as mentioned in this thead.