Which should I follow?
You're going to be scanning your website, what... Annually? Quarterly? Monthly, weekly? And that'll show up on all those scans. Until you tell your scanner to skip that check, or to make an exception... at which point it'll sit there until you have a third party run a scan for you, or one of your partners scans your site and puts it under your nose as "How can you leave such a basic checklist item unaddressed?
Conversely, you're going to be scanned for compliance to the HTTP spec... never. You'll need to actually use TRACE... pretty much never. The practical effect on interoperability with your clients will be... nada.
I understand and agree with @Tom-Leek's point that it's not much of a security issue. However, I disagree in that the downside of disabling is miniscule, and the upside of disabling it is to avoid a lot of annoyance that would otherwise end up in your lap.
I'm not the only one to think that way - which is why Apache added a directive in 1.3.34 and 2.0.55 to simply shut TRACE off:
TraceEnable Off
Here's a good page discussing it, including manual test steps.