When writing XML for large numbers of import articles, I'd typically recommend working with a temporary install in order to get the XML right, then start with a clean install for the production import. This is much easier to clean up.
However, for the situation you describe, it should be possible to clean up with a PHP hack. You need a way of determining all the valid article IDs that you want to keep; then use $articleDao->getArticlesByJournalId(...) to loop through all articles and delete them if they're not in that list using $articleDao->deleteArticle(...). Of course, do a full backup before attempting this.
Public Knowledge Project Team
Don't miss the First International PKP Scholarly Publishing Conference
July 11 - 13, 2007, Vancouver, BC, Canada