I am trying to simulate a crystal filter using measured S parameters. There are a couple points in the response that indicate that the passive circuit has gain. The points are interpolations between the points in the S parameter file. All points not in the sharp peaks agree with measured data on actual hardware. I have seen this in other programs in the past when modeling high Q circuits. It is usually a numerical accruacy issue due to rounding. Is there any way to adjust the numerical accuracy or anything else in the simulation to resolve this?