My system to date has been very paleolithic. I keep a spreadsheet with all my bullet points, categorized by job and mapped to different Skills and Experiences (i.e., the things to be demonstrated via the bullet points). When compiling a resume, I cross-reference by job description and its keywords, skills/experiences to be shown, etc.
Then I sort of treat the process like a compilation of overlapping sets, i.e., "The job requirements are A, B, and C. I want to show that I've done X, which demonstrates A and B. Then I want to show that I've done Y, which demonstrates B. But maybe Z demonstrates B and C in a more effective way, so I'll use X and Z as my bullets." There is definitely an easier and more automated way to do all this, perhaps with databases, but I still value the organic process of poring over everything myself and gut-checking if the overall narrative flow makes sense.
As you might imagine, I played too many RPGs as a kid.
You can output various formats (HTML for web posting, PDF for distribution/printing, .DOC for recruiter's resume-uptake systems (many are based on MS Word / VBA macros). This is also handy for stripping your personal and/or contact information from resumes posted online (to keep annoying contacts down).
Viewed from outside the box: this is a classic example of a clever hack to work around the wrong part of the problem.
Fortunately that tedium is usually brief and infrequent. It's also not current.
And as you said, you want to make sure the overall narrative flow as well as the overall design of the document makes sense for the job.
It's really a cool library; at one point I was going to build a job site to help people build xml-based resumes using this library.
Of course, an army of lawyers put that together - is there an equivalent group to do that for technical resumes? A resume coach at the VFW wouldn't cut it...