Question about PKI-based non-deniable authentication by Canetti et al

In Universally Composable Authentication and Key-Exchange with Global PKI, Canetti et al. proposed a non-deniable authentication functionality as well as a protocol that GUC-emulates it.

My question is that, in the functionality $$\mathcal{F}_{cert-auth}$$, it seems impossible for a corrupted sender $$S$$ to change the identity of the receiver $$R$$ using the command $$(\textsf{Corrupt-send}, ...)$$, while the change of the message to be delivered is allowed. However, in real world, a corrupted sender $$S$$ does have the ability to change the receiver as long as its message has not been delivered until corruption. So how does the simulator work in this case?