This can easily be reproduced by placing something on an analog track for playback through an interface. The corresponding interface output should then be routed to one of the interface’s inputs and punch recorded to another track. If adjustment for record latency was disabled, the recorded track should be delayed by the Round-Trip Latency (RTL) of the interface (the sum of its input and output latencies).
As demonstrated below, a test signal was placed on 4 tracks for playback thru 4 interface outputs. Each of these 4 outputs was externally connected to 4 interface inputs and punch recoded onto 4 other tracks. If Cubase had not adjusted for the record latency on these tracks, they all would have been delayed by ~9.33 ms – clearly, they were not. As expected, deactivating “ASIO-Guard” and/or changing its level did not have a significant effect (only ~0.5ms). The absence of the checkmark box did not prevent numbers in the “Record Shift” from impacting the adjustment either.
The checkbox doesn’t seem to do anything!
download/file.php?mode=view&id=19466
Note: I reported this issue previously, apparently without sufficient detail to prevent it from being placed in the “miscellaneous” file -sorry! I’d be happy to remove it if desired/required (assuming I can).
March 19, 2017 Update:
More details on the measurement procedure and the test wav file used can be downloaded from https://www.gearslutz.com/board/music-c ... 5-9-a.html

