hi egliu,
I agree with you that if couplings can be taken into account the fabricated antenna would be much more similar to the synthesized one.
However take into account the following
1) an antenna design with low sidelobes requires many elements in the array, and if the coupling terms are considered, the problem maybe almost imposible to solve (just in terms of computing resources)
2) in a low sidelobe antenna the excitation errors/uncertainties maybe much more important than coupling terms
3) ideally the synthesis should be done with embedded element patterns (measured to be realistic), which already account for the effect of the coupling, estructure, edge effects,..., although unfortunately implies that you have to build the antenna before
4) many text books (antenna handbook by Lo, microwave scanning antennas by Hansen,...) already include element excitations for low sidelobe array which do not account for coupling and are good starting points for the synthesis
I agree with you that it is better include the couplings, but there are many more effects that should be also taken into account that it is almost impossible to cope all of them. I'm afraid that we still have to wait for a software tool which definitely solves all our problems.