On pg. 357 I give the proto-form of iuxtā as *iugistā. The zero-grade u with a short vowel goes
against de Vaan 2008:318 and Watkins 1975:530 who specifically notes "a ma connaissance aucune évidence ne nous empêche de postuler un ū longe." But the evidence of Old French, which has joste 'next to' and the derived verb joster 'put side by side, joust' points clearly to a short vowel. See the Dictionnaire étymologique de l'ancien français s.v. Forms with u in the various Romance languages are secondary. This fact has interesting consequences for the morphological analysis of this form. It makes connection with the neuter s-stem iūgera even more difficult, but it also doesn't exactly favor Watkins's deverbal superlative analysis since those forms too tend to have full-grades.
In any case connection with the root of iungō is unavoidable and this means that the g of iug- has been devoiced before a voiceless consonant without the operation of Lachmann's Law. The absence of lengthening—cf. fūsus from *fud-to- with LL—might be explained in a number of different ways. It might be the case that LL didn't apply to secondary sequences arising by syncope (hence maximus, despite its lone apex, really doesn't have a long vowel). Or it might be the case that it did apply to such sequences but only if the voiced stop was temporarily reintroduced on the basis of related forms. In the case of iuxtā the derivational connection with iungō was obscure enough that this didn't happen.