I made a COMSOL simulation of our wafer (75 mm with flats, 1 mm thick) with a 1 micron thick coating (Tantala), and computed the dilution factor (E_coating / E_total). The result is shown in the plot below:
The dilution factor is slighly mode dependent, around a value of 5.7e-3.
The Q we measured on the latest two annealed wafers are in the range of 5e6 - 10e6 for the good modes, meaning that the total loss angle (subtrate, surface and edge combined) is 1e-7 - 2e-7.
Assuming an undoped tantala coating with loss angle of 4e-4 (http://authors.library.caltech.edu/55765/2/1501.06371.pdf), the disk loss angle after coating will be 2.2e-6, a factor 5 to 10 higher than our uncoated and annealed wafers.
So we can use the wafers as they are for our measurements.