My recommendation is to change the envelope sender on the auto-reply sent at step (2) of your flow. It is this envelope sender address, which will receive any error messages at step (3).
You can use an empty envelope sender at step (2). That will prevent the error message being generated at step (3). In the SMTP session it would look like this:
MAIL From:<>
Alternatively you can create a separate address for using as the envelope sender at step (2). Then the error message will still be generated at step (3). The error message generated at that step will have empty envelope sender and a destination address matching the envelope sender, you used in step (2). By using a separate address for this purpose, you can treat it differently and avoid step (4).
You do not have to change the From:
address inside the email headers at step (2).
Whatever you do I also strongly recommend that you ensure the mail you send at step (2) contains enough information to uniquely identify which mail at step (1) triggered it. That essentially means include the Message-IDMessage-ID
of (1) somewhere inside the mail generated at (2).