I ran a few more simulations to see how support area would affect the displacement. It turned out that it was not significant, for area = 0.056 x 10^-3, 0.9 x 10^-3 and 5.6 x 10^-3 [inch²]. This is good because we don't have to worry too much about the effective area of the contact points in the simulation. The errors will probably be dominated by other parameters (mostly, support positions). Judging from all the requirements, I think I'm close to making the decision for the support position.

     The plot below shows results from different support angles, (theta = [30,45,60]).

==a few comments about the plot==

• The area of the support points are not significant in the simulation, see green, red, cyan plots.
• For the support angle, it seems that the smaller theta, the less sensitivity to vertical acceleration (blue ,pink, green). However, smaller angle means the support area becomes more vertical. This will cause two issues that I can think of, (1) more force on the spacer which will increase the surface loss, (2) a bit harder to machine the mount. About the loss, I think it is still ok if the normal force from the support to the spacer is less than a factor of 5. This gives us the smallest angle of ~ 12degree, but at this angle, the optimum support point for zero tilt will be very close to the ends of the 1.45" spacer (based on the blue ,pink, green curve) and it will be unsafe to mount the cavities in case they slide out of the supports.

            I think possible choices (considering loss, machining, safety) for the support positions will be some where around 0.7-1.2 inch along the beam line, and the angle will be ~ 12 - 30 degree. I'll run more simulation to see if I can find it.

==Note==

1. I reproduced the result for 8" cavity with wire support and got the same result as before (dL/L)/(acceleration) ~ 2 x 10^-10 (1.7x10^-10, to be more exact), but the tilt is 9x10^-9 rad which is a lot larger than the results from the short cavity. So I think my simulation for short cavity is fine as I can reproduce the same result.
2. I tried to constrained the support area in vertical direction only, but the simulation failed. I think this is because of the bending of the spacer surface. If you let it slide horizontally, the surface will tilt a bit as well and cannot be kept fix in vertical direction. Then for the constrain in all directions, I don't think there will be a solution for zero coupling for the short cavity.