The binaryfile.ini/pdfhandlersettings/textextractiontool documentation page indicates that the pdftotext linux binary file can be used to index PDF documents.

That's not true. eZ Publish must use the pstotext binary.