I understand your intention however you cannot do something that violates physics or math. farfieldvector3d is used in the example file for the whole lens, non-periodic structure.
farfieldexact3d cannot be used for periodic structure, since it does not count for the periodic contribution, and it requires the field intensity at the monitor edges be close to zero, if not exactly zero. This is its theoretical base.
Again, you can either simulate the whole lens with farfieldexact3d, or simulate the unit cell with periodic boundary and plane wave with summing method. Periodic simulation results can only use grating analysis, which is farfield. Then it needs stitching back to the focal plane.