Why do users reply to messages clearly marked as do not reply

In a web app, we generate a lot of emails including reminders, requests for approvals, login details etc. They all come from a real email address to help with deliverability etc and they all quite clearly state:

*** This is an automatically generated email, please do not reply to this message ***

For clarity here is an example of an email as there seems to be some confusion on the type of message being sent and the suitability of "no reply" emails especially as all the requests require the user to login, check the item and then action something:

John,

Michelle has created a new requisition for 6 widgets which needs your approval - please login and process this requisition

https://www.domain.com/app/requisitions/123/process

    *** This is an automatically generated email, please do not reply to this message ***

And an example of the messages we get back:

Michelle,

Why have you ordered 6 widgets? I thought you only wanted 4?

John

The message is in plain text rather than HTML (again to aid deliverability etc) so there is no other easy way to highlight the line - the emails typically only contain three or four lines and this one is at the bottom with white space above and below so it is not lost in amongst lots of other lines.

Despite this, we get numerous people replying to the emails either saying things like "approved" rather than logging in and actually approving the request or they ask questions etc - these all come to us, of course, but they are actually intended for the person who generated the message in the application so each time we have to find that address manually and forward it on.

Why do users ignore this and reply to an automated message? What clever ways are there which we could use to overcome this?

Edit I have no idea why everyone is assuming this comes from a no-reply@ email address - this is a proper email address with a mail account etc