You might want to try making your areal limitation dependent on magnitude. For example
a formula like :
Limitation = 3*((maximum likely magnitude -2)+1)
would allow a distance culling of anything more than 3 times the stated distance
for events of Ml 2, 4 times for Ml 3, 5 times for Ml4 and 6 times for Ml5 and 7 times
for Ml 6 etc.
A Ml 6 will likely be felt over an area of several hundred km so using a 7 times
35 km - if that is the forecast limit - seems not unreasonable.
This would also require smaller earthquakes to be more precisely located
which seems intuitively correct to me.