You might be able to export the target graphical data to a text file of some kind. Once in a text file, you could use a Data Access Object(?) to look into the file and pull out the performance characteristics versus frequency (i.e. Insertion loss vs. freq). If you can get the DAO to give you the same plot (say plot it using ports 3 & 4), and you can plot your desired "matched circuit" response using ports 1 & 2, then your goal equation would be to make dB(S21)-dB(S43) = 0. (well, dB(S21)-dB(S43) Inside -0.1 to 0.1 dB, or similar)
That's what I can think of from a high-level view... the exact implementation isn't entirely clear, but I think something like that may be possible.