Yes, the "phase ambiguity" in the Witten's paper is the so-called _framing anomaly_ ; The Witten-Reshetikhin-Turaev invariant gains an extra phase factor once you change the framing of a 3-manifold, but this problem is fixed by using the canonical 2-framing.
It seems that in the paper you referenced, the $\theta$ map is a Dehn twist, and the "anomaly" is a sort of framing anomaly in this sense.