I don’t know about doing it “right” but Scrivener is the best thing for scientific papers since the invention of academic prose (maybe better, on second thought).
First, you have to disabuse yourself of the idea that scrivener does form. It doesn’t, it does content. So, any tables, footnotes, etc, will have to be added later. Having said that, it’s useful to keep in mind that there are two forms of scientific footnoting: bibliographical (history, PolSci, IR, most of the hard sciences) in which your footnotes include bibliographical information (and the awkward op cit, ibid, and their relatives), and substantive, in which references are embedded in the body of the paper, there is a bibliography at the end of the paper, and footnotes are kept for substantive comments (anthropology, sociology, psychology). I prefer the latter, it has to be said. I find Scrivener works better on substantive footnotes, because it does not handle footnotes too well, or, at least in the same way a dedicated word processor does.
All the above was really an aside. I think ThomasA’s question is really about workflow. Since I have now written at least two major (90-100pg) reports and several articles as well as parts of two novels on Scrivener, I have some experience of how to make the work flow easier.
a. start with your template (or just start a project).
b. I usually collect and import all (or as much as possible) of the relevant literature if possible in electronic format. For example, every time I read an article, I make notes, these, as well as any pdfs are added to the research binder, which I subdivide into relevant themes.
c. write an outline (I use a word processor for that, or simple write it out as a flat text file in Scrivener). Import the outline so that each chapter is a folder, and each section is a document within that folder. The thing to keep in mind is that you can reverse b. and c. if you prefer, AND given the flexibility of Scrivener, you can change the order, merge, delete, etc with no problem. However you do it, it helps to eliminate duplication. If you do the outline part first, then you can make the folders in the research binder the same name as your outline, but I don’t like doing that since of course you use references multiple times.
d. Once everything is set up (takes from 30 minutes to a few hours) you can then write each section. I have started color marking each section as “First draft” “2nd draft”, “Revision after review by outsiders”, etc. You can merge, split, promote, delete sections as you will and as the ms develops. The trick here is LABEL EACH SECTION FULLY. Think of it as a container: the parsimonious way to write scientifically is to ensure that each container contains one ‘thing’, that is, major subject of interest. Since I use MLA/Chicago referencing, I also have a file at the end called REFERENCES (since I use Bookends, which plays very nicely with Scrivener, this is not strictly necessary, but I am old fashioned).
e. Once the MS is to your satisfaction, compile and export it to your word processor. This is where you do the formatting: tables, figures and captions, headings, TOC, etc.
I find (and my experience includes six academic books, over ten novels, and numerous papers) that working that way cuts around 40% of the writing time of a scholarly work. To add to that, I have the feeling (though can’t substantiate it) that it also improves the quality: the references are there for immediate consultation, arguments can be structured in a logical flow, and redundancies are eliminated.
Have fun
per ardua ad astra