With optical techniques today, we often have to use multiple exposure passes to get the features defined in just the right way. This is obviously extremely challenging, especially as you go beyond double patterning into 4x+. This dramatically increases the costs since you are tying up the most expensive tools with all this extra work.
I think the end game with NIL is some hierarchical approach where you create a master stamp (one per product) out of the hardest material, use this to create an intermediate stamp (one per line), and then the intermediate stamp is used to create the working stamp (one per tool).