The aim of this paper is twofold. First, it studies the epenthesis of [I] in Tunisian Arabic (TA) in different environments (C+CC and V:C+C), where it focuses on the way the exact site of the epenthetic vowel is determined (e.g. C+CC > CvCC rather than CCvC). It shows that the reason why a CCC cluster is broken into CvCC rather than CCvC owes to the preference of TA to violate juncture contiguity instead of domain contiguity. This discussion is fully set within Optimality Theory and related sub-theories such as Generalized Alignment and Correspondence Theory. Second, the paper argues that the syllable formed by the epenthetic vowel (CvC) should be added to the TA syllable inventory as a separate syllable type. The evidence comes from stress assignment and shows that the syllable, despite its having a CVC shape, has to be considered as light. This subsequent discussion is set within moraic theory and adopts the technique of mora-sharing in order to satisfy both requirements of monomoraic weight and coda moraicity following the principle of Weight-by-Position. By doing so, it solves the long-standing problem of how to interpret the stress shift in derivatives such as ["tik.tib] (you write) and [tik."tib] (it was written).