Fwd: Is this a Saxon Bug ?

Hello, my problem is very simple though it seems to be unsolvable.
Here's a source file and a stylesheet. In the sourcefile, all
caracters are UTF8 (actually Braille characters). See my XSL sheet ?
Could it be simpler ? I plan to do more things with it but there's a
bug(?) already. This problem happens with some documents, I'm unable
to find out what produces it.

I ran the xsl sheet and this source document with the last SaxonB and
the new SaxonHE, both results are the same. See the difference on
lines 76-79 ? Does anyone know a workaround ? I tried to split the big
text node, it still doesn't work.

The 3 files (input output and xsl) are here :
http://fschwebel.free.fr/input_output_xsl.zip

Thanks for any help,
Regards
Frederic Schwebel
http://natbraille.free.fr/