S/N filtering using ChromaTOF software for GCxGC MS

When I process my GCxGC-MS data using Pegasus ChromaTOF and set the signal to noise value (S/N) to e.g. 50, the software still returns selected peaks below 50 in the peak table. In my mind, If I set the minimal S/N value to 50, I expect to get rid of all peaks below that threshold value, right? What have I missed?

Best regards,

Tim