Thanks Fletch, but you'll see that my second example in the root node already uses pdftotext, and it is dropping the first character on many lines. Yet xpdf displays the PDF correctly!