Converting documents from Word to ODF or vice-versa is full of issues as no conversion here is 1-to-1 (same applies for conversions of e.g. old MS Word formats to newer ones - don't know for various ODF implementations).
If you don't collaborate on the document writing, IMHO pdf is the best representation for both of the formats and all those offices that require ODFs instead of MS Word do accept pdf files as well.
Creating a pdf from a MS Word / ODF file is a matter of seconds to 1-2 minutes for very large documents.
I usually just decide between plain-text, HTML, Google Docs and LaTeX/PDF, going up the ladder as the complexity of the document requires, or if it needs to be editable.
I hate it when I'm e-mailed a Word document that could easily have been plain text, in which case it would be searchable and more easily found when looking through e-mails manually.
In the end though I too use pdf and make the other people do the hard work for not using Open Office on their end :P
I could save the Google doc back to Word format, but I have no idea how different it will act and look.
ODF is for publishing, i.e. where you can have separate pages, with footers, margins, absolute position inside page (versus whole document), corrections tracking, authorship, figures.
HTML _can_ give you close result, but this is not HTML domain, hence separate standard.
Embedded editable files, for instance.
Or markup for tracking changes.
Tex might make a good starting point for building a system that _did_ do all of that, but nobody seems to be working on that.
I've often wondered why Microsoft chose to embed a weak implementation of version control into their mutable document format instead of doing something like passing around a signed chain of immutable documents.
Also, if you don't trust the other parties, you can use the compare feature an get a diff of the documents. Keeping your own copy is a easy, simple, way to get a correct view of changes to a document.
You can do a mediocre job of inferring the latter from the former, but for 100% support you need to actually put the notes into the file itself.
Any representation that makes people think "this is a paragraph", "this is a title", "this is a list" and then apply styles to each is an improvement over the current "lets vomit text on a page, then tweak individual sections independently" method endorsed by most word processors.
Tools that do this like (Pandoc being the one I like) do a great job of getting people to think of documents as text that gets styled later.