The deposition of beam torque differs from the deposition of beam power in one important way: torque can be coupled to the plasma directly through radial motion of the beam ions, i.e., no collisions required. Any radial motion of the beam ions represents a radial current, which when crossed with the poloidal magnetic field exerts a torque. Goldston  describes the physics clearly. Note that because the Fokker-Planck routine in SNAP\ assumes a zero banana-width, it completely neglects the torque (it isn't lost, it's collisionally deposited instead).
The case of first-orbit losses during counter injection is an instructive example of the torque which is neglected by SNAP: Consider a counter-injected ion born relatively near the outside of the plasma with a toroidal velocity , with a pitch-angle such that it is banana-trapped on its first poloidal orbit. Assume it leaves the plasma on the co-going side of its first orbit (because its orbit-shift is outward) with a toroidal velocity . Clearly, the beam ion has imparted an angular momentum to the plasma, even though it was lost on its first orbit! This will be represented as a torque in the orbit codes, because the beam ion effectively represents a large outward-directed . This process has practical implications: throughout the region near the plasma edge where counter-born beam ions are lost on their first orbit, there will be effectively no power deposition, but the deposition of beam momentum will be roughly twice the original beam momentum. Consequently, the profile of deposited torque can be much broader than that of power during counter-injection.
The beam torque which isn't coupled to the plasma as torque, and which isn't lost due to charge-exchange of the beam ion, is collisionally deposited to the plasma electrons and ions. These terms are calculated by the Fokker-Planck routine in SNAP.