fetchmail bounces

[Edward]

I haven't worried about these as yet, but it's time I look at some of 
the error messages and clean them up.

This is the scenario (RH9 all updates):

I use sendmail to send mail VIA my ISP mail server (masquerade).
I use fetchmail in a 15 minute CRON job to fetch new mail from my ISP 
mail server.

All mail for my sub domain gets dumped into one account on my ISP, where 
my mail system then distributes it once it hits my server (fetchmail 
multi-drop).

Lately I've been getting these:

----- Transcript of session follows -----
... while talking to mail.m.iinet.net.au.:

 >>>>>> MAIL From:<FETCHMAIL-DAEMON at localhost.localdomain>

<<< 553 sorry, your envelope sender domain must exist (#5.7.1)
501 5.6.0 Data format error
550 5.1.1 <FETCHMAIL-DAEMON at server>... User unknown

I THINK this is what happens when I get SPAM addressed to a wrong user 
name here locally (it only seems to happen with spam). Fetchmail while 
collecting from my ISP, will attempt to bounce the message. This used to 
work, but I think my ISP has become tighter with security, and now all 
domain senders must resolve (scary it didn't matter before hey?).

Anyway, while sending via sendmail, all my addresses (*@localhost) seem 
to be correctly masqueraded to *@my.resolvable.domain. However, it looks 
like when fetchmail bounces a message back via mail.m.iinet.net.au (my 
ISP mail server), it does NOT masquerade the domain, resulting in the 
ISP bouncing fetchmail bounces back to me with the above error.

The questions I have are thus:

1> Is this a sendmail, fetchmail or procmail setting that I can easily 
resolve?

2> Is there a way I can turn off the bounces with sendmail, fetchmail, 
or procmail settings PROVIDED this is netiquette? What I mean is turning 
of the bounces completely will obviously stop bounces back to legitimate 
people who may think they have a correct e-mail address at my sub 
domain, meaning they will never know they have things wrong. What is the 
correct procedure anyway? Bouncing non-existent users or not?

3> Any other ways to stop this happening?

More info easily available (config files etc.), but I'm not going to 
clog the list just yet.

Regards,
Ed.