This is an example on how to use Mason for lumped element matching circuits.  For matching techniques using Constant Q-curves, you can check out my work book... but to be honest the genetic algorithm does an okay job.  R0, L0, and C0 represent the load impedance which is to be matched.  L1, C2, L3, C4, L5, and C6 represent the matching network.  If you want to use fewer elements, just delete the parts (and reconnect) OR change the value of the device from a variable to "1", which will effectively remove them from the circuit.  

To fix a variable so that the optimizer doesn't play with the value, in the variable block change the "<opt min=..." to "<noopt min=...".  

This is a 6GHz - 18GHz two-section Wilkinson with 20dB return loss and 15dB isolation across the band.  Uses lumped resistors but you could substitute real (measured) resistors if you download the Panasonic library on the Downloads page. This file does generate a PADS Layout file.  I talk about the advantages of multi-section Wilkinsons in my work book...

These are several baluns designed for 1GHz; bandwidth is dependent on the circuit. The first is an LC balun, made up of two inductors and two capacitors.  The second is a coaxial balun, made up of a transmission line.  The third is a Marchand Balun, made up of broadside coupled stripline.  The center frequencies are equal, but each has a different bandwidth over which there is good performance.

This is the file I used to generate the stripline T-junction model used in Mason.  The variable block becomes the arguments of the new model, the calculations are performed within the model, and Mason figures out how to transform the circuit into algebra leading up to S-Parameters.  I've also linked the output Matlab file, it shouldn't be too difficult to adapt this to C, either.