You forgot small sample size and small effects with a few of these studies... :-)

Clearly, pedal stroke is all bullshit... it's a fixed system

This discussion may spur me to write a tutorial. Here are the basics:
The simple fact is that there are times when the knee is extending, roughly, 345 to 155. During that time most people produce knee extension power. Later, roughly 155 to 345, the knee is flexing. During that time most people produce knee flexion power. There are times the hip is extending, essentially 0-180. During that time most people produce hip extension power. People also produce hip flexion power when the hip is flexing but it is not a powerful action so it doesn't account for much. The ankle also contributes power when the ankle is extending.
The one subtly is that the muscles that span the knee must transition from an extension action to a flexion action. Just before they switch over in each case there may be a bit of negative joint power. This "pre-activation" is a well known phenomenon in many muscular actions, not just cycling. When you do single leg cycling with no counterweight, the flexion action stays on longer because it has to completely lift the leg past the top. In this case, the pre-activation of the extension action is postponed which compromises power during the subsequent extension action.
So there you have it. The joints do what they do when they can. Not sure what else people think should be going on.

Cheers,
Jim

turn to the upper body for assistance

Most cyclists do this as well and its referred to as "hip transfer power". Its a fairly small term. Elmer et al. 2011 (abstract below) reported that this term accounted for 3-4% of power in the range of 250-400 watts. During all out maximal sprinting with lots of upper body movement the contribution goes up to 8%.
Elmer did not evaluate efficiency but Korff et al 2009 did. They reported that instructions to pedal in circles and to concentrate on the transition phases through top dead center and bottom dead center of the crank cycle "circling" did not significantly change efficiency; in fact it produced a non-significant decrease in efficiency.
One thing to keep in mind when discussing TT power is that it is way way submaximal. On the order of 20 to 25% of maximal muscular power. So the need to "maximize power" is simply moot.
Cheers,
Jim

Med Sci Sports Exerc. 2011 Oct;43(10):1940-7. doi: 10.1249/MSS.0b013e31821b00c5.
Joint-specific power production during submaximal and maximal cycling.
Elmer SJ, Barratt PR, Korff T, Martin JC.
Source
Department of Exercise and Sport Science, University of Utah, Salt Lake City, UT 84112-0920, USA. Steve.Elmer@utah.edu
Abstract
Separate authors have reported that knee extension dominates power production during submaximal cycling (SUB(cyc)) and hip extension is the dominant action during maximal cycling (MAX(cyc)). Changes in joint-specific powers across broad ranges of net cycling powers (P(net)) within one group of cyclists have not been reported.
PURPOSE:
Our purpose was to determine the extent to which ankle, knee, and hip joint actions produced power across a range of P(net) . We hypothesized that relative knee extension power would decrease and relative knee flexion and hip extension powers would increase as P(net) increased.
METHODS:
Eleven cyclists performed SUB(cyc) (250, 400, 550, 700, and 850 W) and MAX(cyc) trials at 90 rpm. Joint-specific powers were calculated and averaged over complete pedal revolutions and over extension and flexion phases. Portions of the cycle spent in extension (duty cycle) were determined for the whole leg and ankle, knee, and hip joints. Relationships of relative joint-specific powers with P(net) were assessed with linear regression analyses.
RESULTS:
Absolute ankle, knee, and hip joint-specific powers increased as P(net) increased. Relative knee extension power decreased (r(2) = 0.88, P = 0.01) and knee flexion power increased (r(2) = 0.98, P < 0.001) as P(net) increased. Relative hip extension power was constant across all P(net) . Whole-leg and ankle, knee, and hip joint duty cycle values were greater for MAX(cyc) than for SUB(cyc).
CONCLUSIONS:
Our data demonstrate that 1) absolute ankle, knee, and hip joint-specific powers substantially increase as a function of increased P(net) , 2) hip extension was the dominant power-producing action during SUB(cyc) and MAX(cyc), 3) knee flexion power becomes relatively more important during high-intensity cycling, and 4) increased duty cycle values represent an important strategy to increase maximum power.

I can apply continuous max torque through 12, 1 ,2, and 3 o'c, returning to normal torque through 4 and 5.

Please stop by my lab at University of Utah any time you're in the area so we can document your miraculous pedaling technique.
EDIT: I forgot that we've had this conversation before. You're in UK right? I offered to set up testing for you at Tom Korff's lab at Brunel University in London but you said you don't ever get to London. I could also ask Paul Barratt to have look. He's the biomechanist for British Cycling in Manchester. Maybe that's closer to you. I'm sure they'd like see your technique in action as well.
Cheers,
Jim