It is not necessary.

1) if you have only one XTC, you do not need to switch on delay compensation, as it will delay the entire signal by 21 samples with no benefit.
2) 12Mic's converters are faster than the converters of the XTC. So a delay compensation for devices with different A/D converters would not result in analog impulses being aligned in the recording.
3) The 12Mic has delay compensation for AVB, raging from a few samples to almost 100 samples (for single speed), settable in nanoseconds (!) with sub-sample accuracy per outgoing stream. This is called presentation time offset and is an integral part of a real deterministic network like AVB.

The latter means that across an AVB network, you can configure both the signals
- of an XTC, which is converted to AVB in the 12Mic, and then transported on an 8 channel stream,
- and the signals of the 12Mic itself, transported on a different stream with a higher presentation time offset,
to be 100% sample aligned at the receiver, including any differences resulting from different converter speed, and compensating the sub-sample offsets in the converters (for example, the XTC needs 12.6 samples, the 12Mic 5.0 samples to convert an analog signal).

...if you really need it. Historically, delay compensation was necessary when microphone arrays (stereo, quadrophonic, ambisonic etc.) were spread across different preamps, because the daisy-chain of two MADI devices would create a delay (of the extremely short duration that sound needs to travel 3 cm in air!). With 12 channels in the 12Mic, you will usually be able to connect all signals to the 12Mic or the XTC. In the rare cases where you need sample accuracy between two devices, the resulting offset between the signals (with Auto CA switched off) can be minimized if you change the order - 12Mic first, then XTC. For full delay compensation, either do the delay compensation in software (within the channel strips) or use AVB instead.