I have taken a look to the two days of data with the 6MHz on also in Dark Fringe. We let the 6MHz at 0dBm of modulatio depth.
First interesting thing is that for B2 6MHz error signal MICH and SSFS are not separated exactly by 90deg but by 50deg. (for B4 56MHz they are separated by 85deg). However, the loss of SNR for MICH if we use the demodulation phase optimal for the SSFS is only of 10%. If we compare the SNR of both error signals we have (see Figure 1 and 2):
B2 6MHz | B4 56MHz | |
SSFS | 156.4 | 666.7 |
MICH | 78.9 | 355.6 |
The difference is of a factor 4, which means that both error signals are comparable for a modulation depth of 12dBm. Notice that these measurements were done for 3mW in Dark Fringe, so there were non-linearities on the RF channel at the moment of the measurement.