Actually the efficiency is the same because the horizontal orientation would be used only used for the actual takeoff, occuring at low speeds, not the rest of the flight. It's more a structural question, and whether the engines can re-orient 90 degrees without adding too much extra weight.
For military VTOL aircraft, tail-sitters proved too tricky and were quickly abandoned in favor of re-directing the thrust 90 degrees. Even though that requires a vast increase in complexity, the increased robustness and simplified operations more than compensated for the decrease in ultimate performance. The same advantages also saw tilt-rotors succeed where propeller-driven tail-sitters failed. If nothing else, landing a spent re-usable stage horizontally, where the motors only have to hover a nearly empty stage, might be a whole lot easier than trying to get one to tail-sit, along with the problem that once a stage has landed vertically, you have to either lay it horizontally anyway or move it vertically to a building so you can rebuild the stack.
Obviously horizontal launch would cut into the ultimate performance, but since our spaceflight problem is simplicity, robustness, and cost of a system, perhaps it would prove a better, more affordable long-term method, with faster turn-arounds and decreased support and infrastructure requirements.