Thank you for your helpful replies!
I restarted from scratch because I could not figure it out. Everything seems to be working fine now, and there is no more issue with the anti-correlations. I was trying to apply different smoothing operations to see how these affected the results, so something might have gotten confused there.
I will say that I am using the newest version of CONN, with Percent Signal Change, and the data were preprocessed outside of CONN—except for smoothing and band-pass filtering, which I implemented in CONN. So I guess along the way something went awry, but like I said, I restarted the CONN steps from scratch with a new project and all seems to work fine now. If it would be helpful to you I can try to dig around more and see what I did wrong the first go around.
Thanks again for your quick assistance,
Michael