I had the same problem and after extensively exploring different solutions, workflows and softwares, I found Abbyy PDF transformer (ver. 3) to be the most suitable and practical.
This is how I do it:
1. If necessary I "pre-crop" the PDF with Adobe Acrobat (take off useless header, footers, page numbers etc.)
2. pass it through Abbyy PDF transformer to get MS Word file (if necessary fix the image/text areas to get figures/charts right)
3. convert it to HTML (with Word, simple "filtered" HTML mode)
4. convert to MOBI with Mobipocket or Kindlegen
P.S. Unfortunately, neither Acribat (at step 1) nor ABBYY transformer (at step 2) and MS Word (step 3) are free.
Last edited by johndoesecond; 06-03-2010 at 10:45 AM.