What happens is that AFNI (the interactive program) only allows one anatomical dataset per directory (session) to have a completed set of Talairach markers -- once that set of markers is completed, the markers in other datasets are expunged. So you have to operate on the second collection of datasets in a separate directory (that's why I called 'directory' as 'session' in the AFNI interface).

Since you probably had the second anatomical in the directory with the first, the second one lost its markers when the first one's markers were completed. So after you move the second session's data to a new directory, you have to do what Daniel suggested and and the (uncompleted) marker set back to the header and proceed from there.

An alternative is to use the @auto_tlrc script, which doesn't use the markers at all.