You forgot the Python attachment, however I think the explanation is not so difficult. As your transmission spectrum shows, the band gap you mention is the zero-bias band gap; at finite bias, electrostatic effects increase the effective band gap. Provided your calculation is otherwise correctly set up, this is not too surprising.
This, in fact, demonstrates a truly important point: it is very dangerous to draw conclusions of finite-bias behavior from zero-bias properties, as the bias itself influences the electronic structure and thus influences the transport properties. The "naive linear response assumption" which one can find in some articles is that T(E) is independent of the bias, but this is very rarely true.
