A fraction of an industrial machine makes zero napkins.
Your analysis is spot on, the problem can be solved more efficiently with a large scale production. The capital expenditure cost however is similar (43M vs 50M). You point out they huge disparity in labor costs $25M vs .09M but labor is 'free' in India (which is to say there is so much of it available that the price is a lot less than elsewhere).
The incremental solution here wins because a small increment in cash instantly starts feeding the supply of napkins. The 'big machine' solution loses because it takes a huge investment to get to the point where you can make the napkins and nobody is willing to fund that.
There is another problem which are transportation costs. The infrastructure in India is, by all reports, spotty in the rural areas. By creating the manufacturing in the towns themselves you mitigate the transportation problems. That would not be possible with a centralized industrial machine. Further the folks who 'make' the napkins can also sell them so you have a personal relationship with someone in the town.
The incremental nature, the fact that India is labor rich and infrastructure poor, and the comparable capital costs make this an extraordinarily good solution to the problem.