Hmm.. Do you have the rt:any as a starter response on every trial? If so, that might explain it because I believe the "correct" column may refer to the first RT on any given trial. If you have multiple RTs within a single trial then any others need to be defined as correct/incorrect during subsequent analysis. If that doesn't seem applicable, please post or send me a copy of your input file, and I'll try giving it a run here and see if I can replicate the result.