Z parameters - Mason's invariant varies!

Hi

Adding the inductor at Source, the Z11  would become Z11+jwL and hence real(Z11) should remain unchanged. However, using SP analysis (Virtuoso - ADE L), by changing the inductor at Source, real(Z11) changes which results in variation of  U (Mason's invariant which is supposed not to change with lossless reciprocal embedding).

I would appreciate if some one can help me understand why real(Z11) changes when I change the inductor in Source.

Thanks,

HKH