Reading the transiesta manual, I understand that it would be sufficient to optimize the structure within siesta and then use the result for a transiesta run.
However, I have also read that it would be necessary to re-run the optimization with transiesta if the forces obtained from the calculation at zero bias are particularly large. On the other hand, the forces shouldn't be trusted near the electrodes. So, what could be a threshold for forces in a transiesta run? Should one consider, for the latter, only the force on atoms far enough from the electrodes?