Search for "ndr" on the page I posted the link to above and you get over 50 references. But in general, ATK doesn't make any assumption of linearity, so any calculation for a device you make with ATK will be non-linear. Then of course if the transmission happens to be trivially constant as a function of energy and bias independent, you would get a linear I-V, but that's just by accident, not by assumption in the calculation model.