This sounds as if there is already text on the document.
A (albeit lame) workaround would be exporting as TIFF, remake the PDF and then run OCR again.
If the text is along the borders, you can also crop the page(s) so that the text gets cropped off. Then run OCR, and uncrop. Note that cropping is non-destructive, which means that everything outside of the borders will remain in the document.
Hope this can help.
Max Wyss.