Greetings,

  Could you provide some additional information about the
filter? Are you using a model from rfLib to model the filter?
Are you using the Verilog-A laplace function to model the
filter? Are you using s-parameter data, n-port, to model
the filter?

1)

2) In general, s-parameter and ac analysis should not
   match. Ac analysis is peformed in the voltage domain    
   and s-parameters calculated from the normalized power
   wave response (forgot the exact term)
 
    Circuit designers will probably find the ac characteristics
    easier to understand and RF designers will probably
    prefer s-parameters

    Since phase response is related to group delay, it is
    very important for filters.  The "flatter" the group delay
    the less data is distorted going through the filter.

3) Less return loss is better.

4) Passive filters should have infinite IP3 and P1dB

5) Noise Figure is determined by insertion loss

                                                       Best Regards,

                                                         Sheldon