Dear Felix,
The current will not vanish when you put B=0. You are calculating the contribution of the left lead alone.
The one which vanishes is the measured one at equilibrium. It corresponds to the current you obtain with kwant when you sum the contributions of all the leads and then integrate them over the whole Fermi sea. If one lead is set to a slightly higher potential, then the sum of all the contributions will vanish but a small part of the order ev of that lead, and thus, the current will be proportional to this ev. You can have more details about this in the documentation of Tkwant or the thesis of Joseph Weston for example.
For the sign you are mentioning (between method 1 and 2), it is a global sign, that comes from the way plotter.interpolate returns the result. (usually, the current is calculated on all the hoppings in syst.graph which contains the hoppings (a, b) as well as (b, a))
I hope this helps,
Adel