With respect to what? You have three choices: primary (total winding), secondary (wiper to neutral), and upside-down (wiper to hot).
In any case, taking a consistent comparison, the leakage won't be much worse in one position than any other, though I expect it will be minimal in the middle, and maximal towards the ends (following a gentle parabolic curve inbetween). I would think the difference to be small, perhaps 20%.
(I think the "consistent" method would be, one half (hot to wiper) with respect to the other half (wiper to neutral), adjusted for turns ratio. This uses it as a normal transformer (that happens to have a common point tied between the two windings), so the measured value is reasonable with respect to regular transformers.)
I wouldn't expect it to be very useful at frequencies over a few kHz, if that's what you're thinking..?
As an autoformer, the output current capacity (including surge/fault capacity) and regulation (series equivalent impedance) are quite good. Most pronounced for high ratios (above, say, 25%), you get the divider effect in addition to the transformer effect, or in other words, a lot of output current flows in series through the top half of the winding, as well as being sourced from the bottom half of the winding. Of course, at ~100%, all the current is sourced from the hot wire, and there's no series impedance. (You're still limited in current by the size of the wire, of course.)
Tim