Open the .opf in Notepad++ and use a regular expression to replace the erroneous lines with a blank.
I'm guessing that you have 2 images in each chapter, and they're all linked to the same sources on disk. If you want to avoid having to clean up the opf and can accept a bit of bloat in your epub, then a solution might be to create copies of the images with distinct names for each chapter and link to those in your source documents.
I never used CS3 to create epubs, but I do know a lot of people complained about how buggy its epub export filter was. Unfortunately, even CS5 fails to address this issue properly and creates a copy of each repeated image rather than simply linking to a single version.
|