netx fails to launch UTF-8 JNLP files that begin with a BOM (byte order marker). [ http://en.wikipedia.org/wiki/Byte-order_mark#UTF-8 ]
To reproduce this, take any working UTF-8 JNLP file and use a hex editor to prepend 0xEF,0xBB,0xBF. netx will now complain "Invalid XML document syntax".
The same file works with Oracle's webstart. In my opinion a BOM on UTF-8 is pointless. I'm guessing we unintentionally added it by opening our JNLP in Windows Notepad at some point during development. That said, if you want netx to be ultra-compatible it should skip over the BOM. Your call.
P.S. I have not tested how it handles UTF-16 BOMs. Those are actually useful.