Full Server Monitoring for Hack Mitigation

We've been battling this problem for a couple weeks now, and I'm at a loss.

Somehow, someone is able to drop a file into a directory in one of our websites, make mods, and essentially get viagra and related links into our coldfusion site's master template.

I recently placed a monitor on the directory, just to monitor file changes... given the time frame, someone is actually doing this MANUALLY.  However, the web server log files don't show any url manipulation, the FTP logs have no logins for the time period, and the windows security logs show no abnormal activity for the time period.  I even searched the hard drive for all files that changed on the day the changes are made -- nothing appears out of the ordinary.

All Windows patches are applied, and the CF app is as secure as I can make it.  I need to determine the vector they are using to get in and do this simple hack, but without the log files showing ANYTHING, I'm not at all sure how to proceed.

In the past, these things were typically a CF coding vulnerability (SQL injection, etc.) but this is something different.

Can someone point me to some (hopefully inexpensive!) tools I can use to monitor the entire server and determine HOW they are doing this, after the directory monitor shows me when something has happened?  I don't know where to go from here, and it's very frustrating to think someone has this level of control after the system has been patched diligently, and the CF code doesn't appear to be the vector since at some point one would think the IIS log would show a set of page loads needed to accomplish this attack.

I think I have the Windows event log set to show everything (using local security policy), but I need more info.

Help!  I wish I could assign a zillion points on this one... it's a relatively benign hack and a 15 second fix... but I want it to stop, of course.

Thank you...