Praudhamanorama p156

p156

ādeśapratyayayoḥ (Pā. Sū. 8.3.59). Having in mind that a single genitive case ending is differentiated by difference of domain, he explains: "a substitute and a part of an affix." If, however, a part of a substitute were taken, then there would be a flaw in tisṛṇām and the like. And if [it meant] an affix which is the letter s, then it would not occur in rāmeṣu and the like. Although when interpreted with such divergence a dvandva compound is not proper because of the absence of an intention to express them together, nevertheless it must be understood as Vedic/sūtra-style usage.
[Regarding the phrase] "of that which is open and unvoiced": as for what they say—that the most proximate to the unvoiced aspirate s is only the retroflex ṣ—that is inadequate, because it would overapply to ṭha. And likewise the Prātiśākhya [states]: