Maybe if the romans go beyond the Atlas mountains you could get much more romanized berbers.
There's little doubt Berbers were relatively importantly Romanized (as for litterally every people on the Roman border, safe in Asia, Roman political/cultural influence was decisive into ethnogenesis and identity build-up) and integrated within provincial make-up (some could say it was more the case than with European Barbarians). Berbers went quite close forming a Romano-Barbarian entity (altough possibly more or less confederal) as it was witnessed in western Europe.
On this regard, Justinian reconquest was a radical rupture as it was made as much against Berbers than Vandals, and sanctuarized a more important differenciation (altough less so in the western-southern peoples).
One could propose a TL where Berbers manage to achieve their takeover of post-Imperial Africa, and eventually pulling a Merovingian-like unification of it, ending up with a Romano-Berber culture speaking, at least partially, a Romance language, which would include the coastal and inner regions of Mauretania (coverted by the Aures confederation IOTL)
That said, it would obviously ask a PoD predating what the OP asks for, as it's about a Romance language gradually supplanting Arab, therefore asking for a PoD not before the VIIIth century.
After the arab invasion you could have a different Darija, maghrebi dialect. Darija is mainly a mix of berber and arabic.
IOTL, you did have some relevant influence of Vulgar Latin, with some loanwords (not unlike Latin onto Anglo-Saxon). The problem being that Arabs essentially settled the coastal areas were Romance was essentially spoken (and survived until the XIth century), not only disallowing much geographic isolates (except places as Djerba) but essentially located in Ifriqiya, not Maghreb (where it was even more coastal, and even more tiny)