Can co-authors disagree in their own paper?

It is rather unusual, but it has occurred before. In the paper

Piccione, Michele, and Ariel Rubinstein. "Equilibrium in the Jungle." The Economic Journal 117.522 (2007): 883-896.

each of the authors has their own conclusions, marked "4.1. Concluding Comments by MP" and "4.2. Concluding Comments by AR." It should be noted though that the writing in economics tends to be less structured (no such thing as a "method section") and this is a somewhat unconventional paper to begin with.

This might not perfectly fit your bill, but the paper Trialogue on the number of fundamental constants by Michael J. Duff, Lev B. Okun, and Gabriele Veneziano basically consists of three co-authors disagreeing.

Abstract:

This paper consists of three separate articles on the number of fundamental dimensionful constants in physics. We started our debate in summer 1992 on the terrace of the famous CERN cafeteria. In the summer of 2001 we returned to the subject to find that our views still diverged and decided to explain our current positions. LBO develops the traditional approach with three constants, GV argues in favor of at most two (within superstring theory), while MJD advocates zero.

In theoretical cryptography, we have a major paper by Canetti, Goldreich, and Halevi, showing that in general cryptographic systems which are proven secure in the so-called random oracle model are not necessarily secure in the real world. At the end, they each offer different opinions about what this means regarding the usefulness of ROM proofs. (Basically, Goldreich says they offer no evidence whatsoever of real-world security, while the other two say they're not ideal but better than nothing.)