Losing alignment makes it permanently harder to regain alignment. Do this by instead of using a linear scale where current alignment equals alignment gained minus alignment lost, use a ratio where current alignment is alignment gained divided by alignment lost (probably with some constant factor in there somewhere).