I'm not from RME, but a user. To my knowledge this is a steinberg issue. They don't allow negative delay compensation. In some cases where the path you are using is faster, because of newer external AD/DA converters, the delay is shorter then reported with the internal converters.
If you are using the internal converters the delay should be reported as zero, cause it is. Phasing is then caused by something else. One solution is to also send and return the unprocessed audio through the same converters. Delay will then be identical on both.
Another solution though not to practical in every situation is the record the processed audio and then nudge that track for 0 phasing.