The goal of this paper is to argue that adequate and insightful analyses are indeed possible in OT for a unified account of phonological behaviors of some irregular stems. I bring to light some behaviors of /s/-irregular and /h/-irregular stems illuminating a generalization that they share a certain property in their input forms. I propose that both types of irregular stems have a mora associated to their stem-final /s/ and /h/ in the first place, which contrasts with the /s/-regular stems that do not have such an association. The prediction is therefore made that two seemingly different types of stems with the same input specification will show phonologically the same behavior in some relevant environments. This prediction is dramatically confirmed by the fact that the stem-final /s/ and /h/ in question are deleted when they are placed before a vowel-initial suffix. The analysis carries over to some additional cases where the stems are followed by consonant-initial suffixes and where the stem-final /h/ constitutes a part of consonant cluster in the input. In all cases, I show that the present system I propose captures phonological generalizations in a way that earlier approaches to /s/- and /h/-irregular stems could not. All the constraints to be employed in the analysis are independently-motivated in the literature. And there is little doubt that those constraints employed in this article are required anyway in the grammar of any other language.