Current runs from + to -, and for a negative biased right electrode we will expect a current from left to right. The positive current direction in ATK is left to right, thus we will expect a positive current. For a high bias this is what you see, and the high bias current is the expected current.
I assume that you have the same temperature in the two electrodes, if not, there can be a thermal current from right to left and this can explain the result.
Else the negative current must arise from a negative transmission coefficient. A negative transmission coefficient is not physical and may arise due to numerical errors. If you use the krylov method for calculating the transmission you can in some cases get a tiny negative transmission inside the bandgap. Please let us now if you have negative transmission coefficients, and the size of the negative transmission. We would also be interested to have a copy of your system to check the origin of the negative transmission.
To cure the problem try the recursion or the direct method for calculating the self energy in the transmission spectrum.
