I’m trying to use the automatically generated feature for Devanagari. In the cjct_devanagari_rakar_forms lookup in the cjct table, there are two entries with half-form + rakar => rakar-form (shown marked in image below). I’m not sure if this is correct as it will mean that the string will have [kha ha] halant halant ra. Am I missing something?
As far as I understand this the source string should be ha halant ra halant. After the rkrf feature the string is ha rakar halant and the half feature will produce h rakar as it is ignoring marks (in this case the rakar).
Makes sense 
Will it be possible to get the source string combination that represents the glyphs generated for Devanagari conjuncts? There are three categories: Akhand & Nonlinear, Half conjuncts and Linear Conjuncts.