I'm checking the result for the calculation. I think it is too early to celebrate.
Nic suggested that I should use comsol to estimate the coating Brownian noise. There are a few problems:
I double checked the calculation code. I changed m to 65, and stepsize to w0/4000, the elastic energy (U) calculated is still ~ 1.5e-10 J. It did not change much from my last calculation. However, what I do not understand is the analytical result for half-infinite mirror (as given in SK2009 but different from GWINC), the number does not change that much.
I found a missing 2pi factor that causes the estimated noise to be higher and to match the measurement. I'll check the calculation carefully again, but it might be that it's still not coating noise.