I think it's just the way the algorithm works, and that the % demand you see doesn't correlate to actual number of pax, but instead to some sort of internal value.
The way I could see it working is something like: give each airport a numerical value to indicate how busy it is. So LHR = 1000, something tiny = 50. Say Kabul's is 700. So its number for domestic flights is 259. Random small airport in Afghanistan is 100% domestic, at a value of 150. 259 airport to 150 airport over 200 nm generates 50 pax per day. Whereas for a SH flight, Kabul's number is 350, other airport is also 350, so 350 airport to 350 airport over 500 nm gives 200 pax per day. Because there are only a few domestic airports, but heaps of SH airports, that 37% domestic demand actually translates to only 1% or so of actual pax, and you'll see the same discrepancy in any country that only has a handful of airports, whether it's Dubai, Pointe a Pitre, Maastricht or anywhere else.
I don't know that's the reason, but it seems like a sensible way to do the algorithm to me. It'd explain that stuff with domestic routes, it'd explain why somewhere like Kiev has no viable LH routes despite ~10% LH demand, whereas Tashkent has quite a lot with 15 or 20% LH demand, it'd explain various other stuff.