Claim: L2QBXL does not serve a useful purpose
Mathematically speaking, I think directly evaluating the box locals at the target points should be equivalent to going through the QBX centers. A polynomial is a polynomial--shifting the center doesn't change that. (If you want actual FP identity, I think truncating the box local to the QBX order would deliver that.)
If nothing else, it'll save a step in the GQBX FMM.
Comments?