This was a very hard optimization: I was hoping to get it down to 14 or 15 but since I'm already too optimized to have an "initialization" step, I need to deal with the final diatomic molecule in a special way instead. This was no easy feat, since I couldn't place that one in a nice way relative to where the other five end up as in my previous solution, so I had to somehow normalize it with the other one like it. 

I opted for the air-water one and after a lot of effort managed to find a way of getting things to line up, at the cost of one more instruction than I was hoping for.