That's a fair point. There were certainly small discrepancies when I ran it multiple times, but I'll use the median metric moving forward.