Although I'm a very much newcomer in quantum chemistry, I think what you are telling is right. The approach can be improved by properly tune short and long range DFT and HF coefficients (Iop 3/107 and 3/108 in gaussian) .

With no personal experience in modelling excimers, I can only underline two problems here.

Firstly, if you study emission, then the question arises about relaxation of atomic positions after electron excitation, the same as in predicting luminescence spectra. Even when you somehow know a priori the excimer geometry, the vibrational state is probably not defined, so instantaneous positions of the atoms at the emission moment are unknown still.

Secondly, TDDFT is a single electron method, failing completely for two-electron excitations. Excimer forming is not necessarily related to two-electron excitations, but quasidegeneration of energy levels of the complex is usual. There is no guarantee that any single electron method can cope with that quasidegeneration correctly, not to mention the nonadiabaticity here. So, TDDFT may be a starting point for rough estimation. Then, if SAC-CI and EOM-CCSD from Gaussian are not reliable due to the system size, I would consider an MC SCF wavefunction, for which the transition moment between any two states can be calculated in GAMESS US, RUNTYP=TRANSITN.

By using TD option you can calculate the time dependent excited state. Other keywords are Singlet which calculates only the single states and Triplets which calculates only the Triplets states. Keyword 50-50 calculates both singlet and triple sates you can get other options from Gaussian 09 help.