I found a small bug when FAR=1 was used in far_threshold, and FRR=0 in frr_threshold. This resulted in a DIR plot that had a random last value (as it accessed memory at an uninitialized location).
This fixes it. There should be no implication for any other plot done so far.