I agree Robin. It sounds like the documented policy should be updated.
If one is going to propose a merge, then they should have done the research to support why they think the profiles match, including which conflicting data should be kept or discarded and why.
Clearly, someone has to do the research and make the decision. And if one has done the work, why not pass it along to help whoever is actually doing the merge? If they haven't done the work, how can they be sure the profiles in question should be merged at all?
Although, as I type this, I can see where maybe someone is asking for someone else to do the research. On the surface, they think its a match, but not sure how to research it further. However, with the current system, proposing a merge doesn't come across as asking for research help (unless they explicitly say so).
I've proposed merges where I didn't realize some of the differences. It wasn't until I actually performed my first merge, that I realized that someone has to decide. I have a few pending merges now that I'm still researching.
The policy should include something to the effect... if there are any differences between the two profiles, please provide necessary sources to support deciding between which differences to keep and which to discard.