Kalin (2020b), building on Kalin (2019, 2020a) and Kalin & Rolle (2020) has defended a strong model of the architecture of word-formation concerning infixation and suppletion, in relation to linearization and phonology. Among others this model makes the following two claims. First that the conditions for selecting an allomorph over another and the conditions for selecting its position if it is an infix, are separate and irreconcilable. And second that allomorph sets share an edge orientation: left-edge infix allomorphs only alternate with prefixes and right-edge infix allomorphs only with suffixes. I will argue that the first point is unsupported by the typological data covered by Kalin & Rolle and that the latter is directly falsified by the agreement affixes of Sáliba, a language of Colombia.