... I find that the two-stage star removal actually works better if there is some overlap between MMT filtering orders.
First removal - small faint stars -- Orders size 2 & 4. Second removal - brighter stars, orders size 4, 8 & 16. The overlap seems necessary because of incomplete capture of particular star sizes between disjoint filter supports. An overlap in the support helps to catch those boundary cases better.
The result is more monotonic erosion with respect to star magnitude, at the slight risk of increased spider vein growth.
