Accuracy of electronic contribution

I am wondering how accurate the electronic contribution to the entropy turns out if the DOS does not match the experimental/actual ones, e.g. if the band gap is predicted to be wrong. In that case, does it even make sense to include the elctronic contribution with the felec module?

Electronic contributions to the free energy tend to only be significant in metals, not in insulators. And there is no bandgap issue then.
However, if the system is a "strongly correlated" system, then DFT even has trouble distinguish metals from insulators and all bets are off (felec would be useless because DFT would be).

I had been giving it some thought again.
Normally, for strongly correlated systems the DFT community would call for DMFT while DFT+U seems a bit controversial depending on who you ask.
Anyway, in the case of ATAT the employment of DFT+U would be not applicable since one cannot set up a vasp.vrap file containing the on-site U value which can be used consistently in all the structures maps/mmaps would generate. Going from structure to structure the U-value would have to be adapted rigorously since it changes with composition and configuration, am I guessing correctly?

The short answer is: I have not implemented the scripts to do that. But it shouldn’t be too difficult for you to give it a try…