Combline Filter Simulation on HFSS

I'm simulating a simpe 3-pole combline filter with circular rods as the resonators. I am tapping the end resonators directly. For ports, I created a cylinder out of aluminium and the outer shell as a cylinder out of vacuum. I then assigned ports for this geometry. The filter is working but not to my expectations. The problem is the my results are varying too much with the length of the cylinders I have used for ports. Insertion loss, resonant frequency is varying with the cylinder length.

I'm including the HFSS file here.

What am I doing wrong here?

Please reply ASAP

You are doing nothing serious wrong but you try to simulate a resonant structure
so you must be more presize with mesh quality. There are couple advices :

1. If your structure has a symmetry plane - define it in HFSS as well
(plane XY - PerfectH for your case) You will win double in the mesh size.

2. Try to eliminate whole metall part if it possible. Use substraction and then assign Finite Conductivity to selected surface.

3. Driven Modal is more acurate than Driven Terminal solution.

4. Add mesh refinement operation from the begining, ie Surface Aproximation
(22.5 deg by default is quit rough), volume refinement (+30.000 element will good), gap between cylinders refinement where you expect field concentration.

5. Set up itteration point near the resonance , so that HFSS will itself refine the mesh in a proper way. May be you will need run it twice or more.

6. See the convergency better than 0.01 at least.

Here is what I obtained with your example: