I was pointed to this thread about the SuperCMIT latency and I'm happy to comment.
When we designed the SuperCMIT in 2010, we opted for a minimum latency for ch1 at higher frequencies and accepted the compromise that the latency got frequency-dependent. In practice we found no real disadvantage (loudspekers also have a frequency-dependent latency), but nevertheless we told our customers to keep it in mind.
It has always been more a theoretic issue, no practical user identified it as a problem. Actually, there has been one sound engineer that wanted to experiment with the SuperCMIT, ch1, in M/S, where you really need exact phase alignment. For this guy we produced the attached compensation FIR filter.
Remember that the normal CMIT signal on ch2 of the SuperCMIT has a normal, frequency-independent latency. So whenever you need to mix a lav with the Super, simply choose ch2!
When using the SuperCMIT, in my guess 90% of the users use ch1, Preset1, 5% use ch1, Preset2 and 5% ch2. Usually ch2 is only recorded as a backup when the situation makes you doubt the algorithm or when you want to offer an option to post.
SCMIT_inv_Groupdelay_FIR.wav