The main game in town for documenting languages in PNG are SIL International . I think they do Lexicostatistics to establish what counts as a language. There is the potential for error, as most data is gathered by missionaries rather than academic linguists; but any differentiation between dialect and language is leaky anyway. From memory, the sources did routinely identify dialects within the claimed languages; so it wasn’t a case of lazily calling every village a different language, and I don’t remember the languages looking particularly close. (Unless they’re Oceanic languages, of course, but that’s a later migration, and not the majority of PNG languages.)