it was stupidly large oversite, until 4.5, where you finally have the same behaviour as transform.
now if only tranformational snaps would respect the manipulator axis constrain state. dang it. to be able to snap w/i a plane, not 3. 'course you can do afore mentioned locking of trans channel for transform nodes, but its of more use w components (and the problem you've noticed - their trans channels dont lock, unless you do some extra poo). perhaps even a constrain to an ortho view pane you were snapping in would be ok? sigh. simple things should have simple solutions .. as they would in their coding implementation no doubt, these are things the users should hammer in the wish lists - oversites/functionality that are easily sorted. it is nice in 4.5 you can retain component spacing when you do snaps, is good. more please!
and more transformational frames, like selecting a component out side your selection to be used as a transformation reference.
so until you lay hands on 4.5, yeah, using clusters/lattices is a method, cumberson, but you can script-automate it easily enuf. then you'll find that the scaling axis isnt oriented how you please, so other stuff (rotate both base and def lattices, or add another tranform oriented as desired) if your polys, suss polyMoveComponet, and scale using channels in channelBox. search the forums and you'll find a bit on this.
g.luck!