The source bandwidth you set is to make sure the wavelength speficied has suffiicent amplitude/power injected.
When you change the source wavelength, the mesh size will change. But I guess you did not change the PML layers. PML absorption depends on the thickness (and PML parameters). SO when mesh size changes please make sure the PML has not the same thickness or relative to the min wavelength.
When the transmission has postive and negative values, it means the simulation terminates with insuffiicent simulation time (did not reach the autoshutoff min, or the autoshutoff min is too large).
I would suggest to use one wavelength range (it seems you use fixed refractive index of 4. it makes things relatively easier), make sure the simulatio time is enough, the autoshutoff min is small enough. You can use a time monitor to check the time signal near the end of the simulation and see if it is small (however the log file gives the averaged result but the time monitor is local); the transmission curve is smooth without non-lasing ripples. Then with proper setting ajustments do other simulation.
BTW: what is the purpose to simulate different bandwidth?