Trackless should be impossible because you can't move any of the accessible atoms off of the lattice defined by the arm, so this should be min cost. The difficulty with this solution is that there is insufficient space to move atoms past each other, so the left over water from the first projection gets in the way. This is made more annoying by the fact that lone atoms cannot reach the bonder, so I take one of the left inputs early to keep a three atom intermediate structure. I swap it with the third left input by bond juggling.