The reduction in the ankle frequency between the 1/f^4 part of the loop and the 1/f part of the loop from 33kHz to 22kHz does not reduce the gain below 10kHz by a factor 5 but by a factor 3. This is the gain reduction is (33kHz/22kHz)^3, the exponent is 3 and 4 because the slope changes from 1/f^4 to 1/f above the ankle (and not from 1/f^4 to a constant).
Figure 1 shows that we see the factor ~3.3 increase in the frequency noise seen on the SSFS and on the IMC reflection due to the reduction of the IMC loop gain below 10kHz. Below 3kHz there is no increase in noise as the SSFS loop has sufficient gain to bring the frequency noise at the output of the IMC below the sensing noise of the SSFS, hence below 3kHz the SSFS sensing noise still dominates the laser frequency noise.