If you are really interested in accounting for two-particle excitations (excitonic effect), there is no way you can do it with ground state density functional theory. You then need to adopt the BSE approach, which is not implemented in ATK. For very accurate calculations of single-particle excitations, the GW approach is an option, but it is not yet in ATK.
However, you can still calculate the single-particle spectrum, using the LDA or GGA functional, as descried in this tutorial
http://docs.quantumwise.com/tutorials/optical/optical.html about absorption spectrum. I think something like this was calculated in one of the papers you have mentioned. I note again that the excitonic shift, which is calculated with the BSE, will be missing in this spectrum. If your material of interest is h-BN, you might get away with the band gap problem of DFT since h-BN has a pretty large gap.
Regarding optical transitions related to the defects, I think you still need to do it in the way Ulrik suggested, see Eq. 3 in that work. It might be useful to have a look at the following paper by Chris G. Van de Walle1 and Jörg Neugebauer [J. Appl. Phys. 95, 3851 (2004)];
http://dx.doi.org/10.1063/1.1682673, especially, Sec. 7D, where they do discuss thermodynamic transition levels versus optical levels, in particular, related to photoluminescence. That review paper appears to provide useful information how experimental measurements for defects can be interpreted from first-principles calculations.
