This is probably crazy, but anyway. There was a discusson on the Halfbakery (http://www.halfbakery.com/idea/swoop_20 ... 1306701675) which led to some interesting discussions on things like the Oberth effect. One particular point that struck me is that, when a rocket is getting near orbital speeds, its fuel contains more kinetic energy than it does chemical energy, which is freaky but true. (1kg of TNT moving at 4km/s has a kinetic energy of 8MJ, and a chemical energy of 4.6MJ)
The bottom line is that a lot of the fuel's energy is wasted, because the exhaust has a high velocity relative to the air around it. This velocity represents wasted kinetic energy, since it is eventually absorbed by the air as the exhaust gases slow down.
If you can arrange for the reaction mass (ie, the exhaust) to have zero velocity relative to the surrounding air, then in theory everything becomes more efficient. However, this also means that your thrust is negligible, if you're exhaust is a gas.
However, you can get around this by (a) starting with the rocket at some reasonable speed (say, 500m/s - it's released from an aircraft or whatever), and then (b) ejecting mass at a velocity which is equal and opposite to that of the rocket, so that the exhaust velocity relative to the atmosphere is zero. Because you're aiming for low exhaust velocities (but increasing in proportion to the velocity of the rocket), it actually makes most sense to eject lumps of stuff (ie "ballast") as the reaction mass, using a relatively small amount of propellant.
I know this all sounds weird. But, if you do the maths, you can get things like 5% of the launch mass delivered to LEO (ie, around 7km/s), using LOX/H2 as the propellant (and not a lot of propellant either). With a solid fuel, you're not much worse off (eg, a 100kg launch-mass giving about 2kg at 6.9km/s, from a starting velocity of 500m/s). Of course the calculations are simplified, but then again so are those for the standard rocket equation, so the comparison is fair. And before you tell me I'm nuts, the maths goes back a long way, and doesn't break any paws of physics.
If you're interested, take a look at the discussion on the Halfbakery and try to figure out where I've gone wrong.