I completely support the idea of applying best practices in both methodologies, at the same time the termonology of enterprise worries me when the attempt to refactoring any portion of subsystems come into place.

Considering that definition of elaboration in RUP looks very idealistic, I would like to point to a few facts that causes RUP's elaboration to win over refactorying. (this is for your point on elaboration for large  scale enterprises which I do not agree with your comment) and this is why:

one is the scalable nature of enterprise systems which requires clear and percise deliverables for each iteration for future added functionalities (new feature). it is required to provide a basis for any future added functioanlity in the system  by focusing on the architecture of the system (again this might sound very idealistic). The other fact is the policy of team rotation which is a reality. This mostly happens in very large organizations with the intention of circulating information among engineers by giving them a big picture of what's going on (some don't like it) or moving teams to new services and replace them by another rotation or co-ops. therefore lack of clear architecute (due to refactorying method) and not having clear and focused documentation for new team will cause cost of time and trouble for new guys to figure out what was going on before. I always say a clear or visualized information (in the form of UML) is much more understandable as oppose to digging into the code. this is why elaboration becomes a key factor in large projects.

you might say refactorying can still produce focused documentations, but I don't agree cause overtime it won't.

on your last point I see we both agree. the decision depends on the size of the projects  and complexity. RUP or any other apprcoah might not be suitable for every process, but an estimate on the project and how it is going to grow surely will give  an understanding as which method is more suitable (which might not neccesarily "seem" to be the proper one at the time). the difference of these two is on how focused they are and depending on the project.

now that we are here (I might sound like a fan) but for large scale  projects I have used RM-ODP a lot and I consider it very successful. RM-ODP may not fall into category of software development  processes as it is more like an architecture technique, but within a bigger scope of RUP it might be usable.(worth checking)