That's probably because if you've told it to stop at a particular place, it needs then to have CDF simInfo for that cell to tell it how to netlist it. Normally CDF is not needed for intermediate levels of hierarchy, because provided the terminal order (etc) is consistent between the instance and the definition, all is OK. If it's become a leaf cell because it's in your stop list, then it needs something to tell it what to netlist.
I can only imagine that the nand gate already has some CDF (although perhaps it is incomplete because of the missing pins), but the cell for i3 doesn't have any CDF.
Setting "schematic" as your stop list is a bit of an odd thing to do, although I presume you have a good reason to do it?