Long-range lattice summation techniques such as the Particle-Mesh Ewald(PME) algorithm for electrostatics have been a revolution to the precision and accuracy of molecular simulations in general. Despite, a performance penalty, few biomolecular simulations today are performed without lattice summation electrostatics. There are increasingly strong arguments for moving in the same direction for Lennard-Jones (LJ) interactions, and we recently made a new fast LJ-PME implementation available in the Gromacs package, where we relied on approximations of the combination rules in reciprocal space to reach high performance. Here, we propose a new way to correct for these approximations that achieve exact treatment of Lorentz-Berthelot combination rules within the cutoff, and only a very small approximation remains outside the cutoff in the reciprocal space component. Not only does this improves accuracy by almost an order of magnitude, but it also achieves absolute biomolecular simulation performance that is an order of magnitude faster than alternatives. The implementation in Gromacs * To whom correspondence should be addressed 1 includes both CPU and GPU acceleration, and combined with improved scaling LJ-PME simulations now provide performance close to truncated potentials, and much higher accuracy.