There is a little bit of luck here. Some places withing a grid capture more squares than other places.
Finer granularity of data would avoid it but then, the number of squares would explode and not be manageable / feasible to deal with.
similar problem with LFPO/LFPG for example. and i didnt look for long. at least make the grid like 1/4 or 1/9 or something, but the mapgrid size is way too big compared to the airport influence circles. if you dont have the population details for each specific little-grid, just take the general one of the current grid for each one, but at the moment, it seems like the location of the airport is way too much connected in relation to the grid arrangement.
also imagine an airport is growing, so the airport influence circle is growing. now if with one increment growth of the circle you increase the number of mapgrids by 50% or 100%, that will be a bit of a mess for the airlines operating that airport.
-------------------
additional questions:
1) which airport is preferred by the passengers of a specific grid?
the closest? or is that also depending on the airport size? (distance to airport in relation to the airport influence circle) [see added graphic]
2) how are passengers of one mapgrid distributed between the different airports?
is it also a matter of frequency? are flights with less than 60 minutes in different airports counted as one flight?
3) why are grids "higher" (north/south direction) when near the poles than they are at the equator?
if i check MPTO, the grid is about squared. for paris it seems rather more of a rectangle with double the size in N/S direction compared to E/W direction.
now i am rather sure, that has to do with the map-projection, but shouldnt the airport influence circle look eliptical in that case as well, when nearer to the poles?