I use Netbeans 6.5, and I've also tried 6.7rc2, on a SunFire V210 and a Ultra 45, both running Solaris 10 and using 
jdk1.6.0_14. Our IT folks have done a bog standard install + installed a few corporate bits and bobs. This
means

pfloyd> locale -a
C
POSIX
iso_8859_1

My project properties all have Encoding as UTF-8. However, selecting any of the
other (many) encodings never seems to change anything.

I could ask for more locale's to be installed. This would be likely to take a
long time.

Here's a quick example. If I create a source file with

// é

[thats a // comment, space and e-acute]

with xemacs [on Solaris or a Linux machine with locales galore] then I end up
with

pfloyd> hexdump accent.c
0000000 2f2f e920 000a
0000005

so the e-acute is being represented by 0xe9

If I open this file and save it with netbeans, it displays the e-accent as a
little box. If I save the file it becomes

pfloyd> hexdump accent.c
0000000 2f2f 3f20 000a
0000005

The 0xe9 has been transformed to 0x3f - the question mark character.

If I write an e-acute with netbeans [compose e '], then it directly displays a
question mark, which it saves to the file.

I'm not particularly bothered about not being able to write non-ascii
characters, but the transformation of existing ones into question marks is a
pain.