It seems to me that an advisor is in a better position to determine "potentially interesting results" than a grad student is; science in the extreme should be i

It seems to me that an advisor is in a better position to determine "potentially interesting results" than a grad student is; science in the extreme should be ideally purely content-based (the proposal alone), but reality is that its not. I imagine you're mostly funding humans (who have provided a rough, or even detailed plan; which will likely be subject to modification or analysis as the project evolves), in which case experience is a strong heuristic. And presumably, if you can't convince any possible advisors that your research is "potentially interesting", it seems unlikely you'd convince the funding department, or even should be able to convince them (who presumably have limited knowledge of the precise area of focus -- I imagine they rely on the deeper specialist input). Of course, if you can convince an advisor that its potentially interesting, there should ideally be nothing stopping further development (eg no one should be claiming that it originated from a student, so it can't possibly be valid -- it's been validated by the advisor, so that's the heuristic to now beat)