An analog sizer's railing gate withholds predicted gain on 47 of 942 simulated circuits, all 47 false positives that SPICE biases and measures fine, because the gate compares the voltage gap against Vdsat, argued to be the wrong tolerance.