Hi,
you should check why the dscf calculation at the last geometry failed. In particular you should check the
geometry and the orbitals (and their energies). It might be that your system dissociates in S1or that it
relaxes along a path where in the ground state a bond will be broken or some other multireference problem is met...
I'm sorry to say that: but excited state calculations with response methods (TDDFT, but also RPA, CIS, CIS(D), CC response, etc.) rely on an appropriate description of the ground state by the underlying ground state method. If this is not (or in geometry optimization: no longer) the case, they can not (or: no longer) be used.
Christof