It also explains the "proximity blast" thing: the beam reaches its target along the most efficient channel, but meets just empty vacuum there and has nowhere to go, so it splashes out from "phased space" into realspace in a fireball.
It doesn't explain how you can "walk" your phaser to the target on occasion, but it does explain why such a thing is done so very rarely: each time you move the tagged end of the phased conduit, you in essence need to create a new conduit, so it's better to just hold on to the one established tag even if you can easily tell that "walking" the beam a few inches to the left would create a deadlier effect. (This especially wrt hand phasers.)
Of course, "moving" and "not moving" are difficult things to define. It seems that it's fine to have your emitter aboard your starship corkscrewing around the target at high speed as long as the phaser connects at a specific spot on the equally corkscrewing target ship. This could be explained simply by saying that dwell time is essential for penetration of defenses, but it would also be consistent with the "tagged conduit end" model, assuming that the phased realm has special properties or "fix points" around starships (not a bad assumption).