I started this as a comment, but it got too long. I have been working for much too long on a multi-extruder printer, and so I understand that the fixed, dual-head designs are somewhat fraught with problems.
I am assuming that this is a mechanism in which the two nozzles are inline in the X-axis, the Y-axis is parallel to the bed and normal to the line between the two nozzles, and the Z-axis is normal to both X and Y (and usually vertical).
When you "bed level", is it clear that both nozzles are "exactly" the same height from the bed everywhere on the bed? If there is anything non-parallel about the movement plane and the bed, it can cause one head to dip lower than the other when over the same position on the bed.
Is twist along the Y-axis of the heads possible? A twist could be because of either looseness or compliance (springiness) in the X-axis. The twist could be induced either through acceleration or by a nozzle being pushed up in reaction to the plastic extrusion against the object.
Is the printing from the active nozzle properly flat, or does it produce verticle bumps? Any vertical ripples will cause any head passing over them to bump against them. Something will need to comply with the force. Either the head will melt through the bump and smear it out a bit, the bed will be pushed downward, or the nozzle pushed upward. If the passive nozzle hits a bump, that compliance will be conveyed to the active nozzle, creating another defect.
Is the passive nozzle cold or hot? It may help to heat the unused nozzle so that it can melt the plastic it touches rather than pushing against it. If this helps at all, it may be necessary to slow down printing so that the passive nozzle has more time to melt through the obstacle.
If you find that the problem is bump defects from the active nozzle, it might help to remove the passive nozzle completely and focus on printing a smooth, even surface with the active nozzle. Debugging that would be easier without the passive nozzle compounding the problem.
