Huhh, can you explain your algorithm for reducing the arrays a bit further?? I see at the moment plenty of possibilities, e.g. I don't see why

1. 3
2. 4
3. 8
4. 17

might not be a valid reduction ...

-- Hofmator