Spectre's pss is not needed. Simply do transient and dft(fft is not used in calculator). Then take the difference of the harmonics. But that is not the true story if you apply the phase shifter signal to a mixer input. Then also the harmonis contribute to the phase performance.
To get the real performance you have to setup a phase shifter plus a down converting mixer. Then measure the phase performance of the downconverted signal.
I am doing this calculation too (but i am working with Hspice) so i do it using ".meas" it is an order that must be configurated, and it is able to give me the difference between the original and the shifted signal, in ascii format. After simulation has finished, i can plot the result in gnuplot.