Hi Tzu
If I read you correctly the methodology you describe, as far as I know, has not yet been proposed by anyone academic. Remember our subject here suffers from a distinct lack of academic interest. The last serious academic treatise on this was carried out more than a century ago by Professor Child. Even that dealt with a very small corpus of ballads and he struggled to come to any sort of concensus/conclusions generally; although we all accept that by and large his tracings of the histories of the individual ballads was exemplary.

Forming the Roud Index has presented problems in this area. Steve still hasn't got to grips with related ballads and hybrids yet. I have made some inroads on this by sorting out the many 'Died for Love' family ballads and allocating new numbers based on this, but then Richie comes along with one of his detailed studies and tells me that even one of these oecotypes should actually be split into 2 separate ballads needing a new number for one of them. New evidence is appearing all the time as more and more collections come online.