You're reading the Nexgen text data via Virtual port in the Magic RDS, is that true?
The setup is according to Help - m4vp.pdf, section 'Redirecting to External Text Sources'
Such feature uses a 'line separator' so two lines are like two different songs sent immediately one after another. Potential additional time separation may not be reliable so it is not used at all.
The line separator and the separator between artist and title may be any in general, they are fully configurable in the Magic RDS. But they can never be the same (or one as part of the other).
What to do next?
Establish two connections: for the Deva (Generic UECP) and for the P164 (P164)
In the Deva connection, enable the RT+ software encoder (Overview - Extensions)
Fill the RT+ template in the external text source, according to final artist/title separator
Optionally, fill the Text replacing rules if necessary.