The Lucene TextContentIndexer class indexes the content of a document twice : first time it indexes the document raw content and second time it indexes content coming from a content extractor.
Created attachment 15530 [details] corrective patch the 'raw content indexing' is commented out as content extractors should do their work right after.