Well, I guess the process is no different than trying to create a graph for, let's say, landing distance. You first figure out what the proper and achievable technique is. You then figure out the baseline results for ISA conditions, MTOW, nil wind etc. And from there you start figuring out what sort of difference different conditions make.
Having seen, and worked with the landing distance graphs in the PA28 POH, I am convinced that it should not be beyond the average PPL to use a more or less similar multi-step graph to figure out the turnback height. But you are right: That graph will be aircraft-specific. Although the research may yield some generic results that would make it easy to adapt the graphs to different aircraft.
And I would not be surprised if the eventual graph contained a "here be dragons" area or something like that. Meaning: Do not go here or there's absolutely no way out. Don't single engine helicopters not have a similar graph for that part of the flight envelope where you cannot survive an engine out (because you're too low and slow for autorotation, but high enough to end up dead)? Pilot training should then simply be about risk mitigation and avoidance of those circumstances.