The question of what you mean by "siphon" effect for the wedge's ability to draw power has arisen. Did you mean that this effect worked for a ship in normal space -- that is, does a ship in n-space draw power from hyper -- or only for ships already in hyper (as in the sail's ability to provide energy from a grav wave interface)? And, if it is operable in n-space, is it also operable inside the hyper limit of a star, or only outside the hyper limit? Is the siphion effect necessary to sustain the wedge, or only to initiate it? And just how much power does it take to accelerate a ship in the Honorverse, anyway? Some folks have been making calculations based on the energy required to accelerate Honorverse ships, and some other folks have been suggesting that without knowing exactly how acceleration is applied in the first place, there's no way to know if the calculators' initial assumptions are sound.
Whenever an impeller drive vessel activates its wedge, whether in normal-space or hyper-space, it "bends" space around itself. Remember that the Manticoran Wormhole Junction itself is essentially a huge gravity anomaly which actually breaches the barrier between normal-space and hyper-space in a fixed, localized phenomenon. It is the "bending" effect on the Alpha wall, for example, which makes it possible to detect and track impeller footprints in real-time at ranges of up to several light-minutes. The same phenomenon operates in hyper- space, except that the hyper-wall being bent isn't necessarily the Alpha wall; it's whatever the next higher hyper-wall may be.
When the wall is bent, energy is siphoned across it from the "higher" hyper band on its other side. Hyper-space is an area of inherently higher energy levels, and the siphon effect could be considered a sort of strictly limited, primitive ancestor of the "core tap" in the Mutineers' Moon Universe. The initial power for the wedge has to come from internal sources -- current generation and stored power. Once the initial energy investment is made, something like 60% of the energy necessary to maintain and power the wedge is drawn through the "siphon" effect. Warshawski sails, because of their interaction with the gravity ways in which they are used, provides a substantially more powerful version of the siphon effect; that's why a ship moving under sail can provide its total energy budget through its sails, whereas a ship moving under impeller wedge cannot.
The siphon effect is operable anywhere, anytime. The ability to raise a wedge, which interacts with the hyper wall, is not the same thing as being able to translate an entire starship across that hyper wall.
The energy required to accelerate under impeller drive is not appreciably greater than the energy required to maintain the wedge in the first place. If you go back and look at the description of how the wedge works, I talked about treating the ship moving under impeller drive as basically being a surfer in the curve of a wave, or perhaps another possible example would be a piece of ice, or a melon seed between two fingers which are perpetually "squeezing" it. There's no reason why an impeller wedge could not the instantaneously accelerated to the speed of light. The problem is that accelerating it to such velocities would instantly kill/destroy any ship "riding" the wedge. Where the power requirement does shift as acceleration rates increase is in the power requirements of the inertial compensator. This comes as the consequence of two separate requirements. First, while any impeller wedge can be accelerated as described above, the intensity of the wedge, the power or strength of the wedge, is directly linked to the depth of the inertial compensator's "sump." The more powerful the wedge, the greater the amount of inertia which can be "dumped" into it. Second, the power load on the inertial compensator climbs as the amount of inertia of being compensated for increases. Thus a superdreadnought accelerating at one hundred gravities would require a great deal more power for its inertial compensator than a LAC accelerating at the same rate.
Missile wedges use a slightly different variant of impeller technology. They don't use inertial compensators at all. Or, rather, the minimal compensator effect needed to permit them to accelerate that such velocities without suffering serious damage is built into the nodes themselves, and not into a separate compensator. There have been efforts -- quite a few of them -- over the centuries to adapt this type of designed to starships, but without success, primarily because the amount of inertia which can be disposed of is insufficient for a manned vessel. Missile components can be engineered to survive the equivalent of thousands of gravities of acceleration; human beings can't. A portion of all of the man-centuries of research which has gone into trying to find some way to adapt missile propulsion systems to manned vessels did find its way into the "beta-squared" nodes of the RMN, but even Manticoran is still hopelessly far away of finding any way to build compensatorless manned vessels.
The fusion of "compensator" capability with the nodes themselves also requires that missile wedges be much more powerful in proportion to the missile's size than an impeller wedge must be in proportion to a manned vessel. (This, by the way, ought to be another clue to exactly what the "siphon" effect is doing, since there's no way in the world even a fusion-powered Manticoran MDM could sustain its wedge [at] such high levels for so long without additional power coming from somewhere.) This design feature is also the primary reason that missile drives cannot be reset or modified once the missile has been launched. It's a fundamental engineering problem which no one has ever been able to overcome, which is why Manticore went to the multi-drive concept in the first place.