Well, I was actually still making up my mind around this. It sounds like a great tool, but I doubt if Sigil would be the best for it.
My biggest questionmark would actually be the first step, from a wordprocessor to 'clean' XML. Clean XML is important, because if it is clean, it is relatively easy to go to anything else again. I agree with Hitch that an ePUB would just be an export/conversion product of that XML, just as e.g. a PDF could be.
The first step however is where the pain is. There are a couple of painpoints there, most of them already mentioned by Hitch.
- Most writers are still stuck in the typewriter age, they are not making use of the actual tooling available (styles, etc)
- Writers don't want to change/learn, it should work as they want (tab or spaces instead of style with indent, enter instead of margins, etc)
- The large number of wordprocessing programs with all different formats
- The garbage these programs spit out as HTML/XML/etc
It will be an almost impossible task to be able to filter/convert the output of all these programs to XML/XHTML while maintaining all the markup and taking the bizar things writers do in their documents into account. I only do it for Word and that is already a nightmare sometimes. Writers still surprise me with their workmethod and output.
There are basically two methods. Either take the native format and convert that into clean XML or use the XML/XHTML output from a wordprocessor program and clean that up if you can. I don't know if you have tried it with Word output and XSLT, but good luck. In all the years it is available, there is not even one good XSLT out there that actually works. And that is just the major wordprocessor.
The ambition is good, but the number of writers that wants to be bothered with this is very slim, especially for novelists.
However, the second part of converting the clean XML to other outputs could be very useful. That being said, XML itself is meaningless without the structure. What structure should be used? XHTML? A kind of LaTeX perhaps?