I was going to suggest the render()
option, with the caveat it's a bloody awful bit of functionality in OpenBDML (or whatever they want to call their version of CFML), and I think should be avoided. I can't see how it doesn't cause the CFML to be recompiled every request, which is a helluva overhead. I think you're better off modifying the mark-up when it gets saved/read to wrap/unwrap it in <cfoutput>
tags, that way the resultant CFML will only be recompiled when the file actually changes. Which will be really a lot less often than it's read.
Prompted by your question, I did a bit of a write up on the CFML compile process, showing why having the <cfoutput>
tags in a separate file doesn't work. Apologies it took so long to write & follow-up with here.