Whilst my thoughts would turn to rigging initially I would also want to conduct a full set of stall checks at altitude including approach configuration replicating that you have set up on the problem turns.
For the left wing to drop it must be losing lift in comparison to the right which could either be early deployment of the right slats, or late deployment of the left. If the rigging is set correctly the only other way of producing this would be out of balance flying, slipping into or skidding out of the turn. Whilst most people regard nose wheel spam cans as not needing rudder I find it quite the opposite. The rudder should be in use in all turns and to assist maintenance of heading in straight flight.
Are you perhaps introducing too much rudder into the turn? Or not using any when it's actually needed? I have a colleague who's concentration and absorbance into the approach see's him at almost full left rudder by the time we get to short final. I could do with a pin to reach over and prod his thigh to release it.
If you have a friendly instructor to hand maybe an hour spent exploring the aircraft and your handling might reveal results. You could count it towards your SEP validation.
Hope you find the answer anyway. Good Luck.