Dave, out of curiosity I created a circuit model for a common-mode choke. It has an inductance, capacitance, and resistance in parallel. I chose the values to mimic attenuation results for a common-mode choke someone sent me a couple years ago. It's shown below. The choke was measured with a DG8SAQ VNA with all four S-parameters. The software that accompanies the VNA can implement what they call scripts. He ran the Y21 script and sent the results. One of the neat features of the Y21 method is that it can calculate the shunt capacitance at each port as well as the DUT response with the capacitance nulled mathematically. Calculated shunt C was 2.8 to 7.3 pF depending on port and frequency. (I don't know why it varies with frequency.) I used his choke to create my choke model. When I added a shunt 2.7 pF to each side of the choke in the model, I saw no difference in choke attenuation to 30 MHz. I thought at least it would be noticeable, but the curves overlap. I tried 27 pF to make sure the model was working, and I could then see a small difference. 2.7 pF is about 2k ohms at 30 MHz. This is comparable to the usual several-k ohm CMC impedance so I expected some effect. You can notice the Y21/S21 difference below for a shunt capacitance roughly double what you measured. The effect isn't great and is of no consequence for choke attenuation, but it would be measurable.
One thing I often forget is that measured capacitance for a test fixture like yours includes the connector capacitance. It is not stray capacitance. It matches the connector inductance to form a 50 ohm impedance. You should subtract the capacitance of a bare connector from the 2.7 pF you measured to get the stray fixture capacitance. There is also capacitive coupling between the choke and the VNA enclosure if it is nearby. That will contribute to shunt capacitance.
Based on my circuit model. I think the capacitance of your fixture should have a negligible effect at HF. It would be a good idea to verify that expectation by measurement.
Brian
