As we've found in the past, using traditional 50/50 nomenclature is incorrect.
I could probably write a thesis on how the 5MT is built and works. I'll try to be brief.
I think the best way to describe this, is to describe the center differential.
The output shaft of the transmission has two parts. There is an outer part, and an inner shaft. The outer part is directly linked with the output of the transmission. The inner part is what drives the front differential.
Now the center diff contains very closely spaced plates surrounded by a really thick, gooey, silicone fluid. This goo has a high CTE (Coefficient of Thermal Expansion). The plates alternate attachment points. Half of them are attached to the diff casing itself, the outer part directly linked with the transmission output. The other plates are attached to a shaft that slides over the pinion shaft, effectively linking them to the front differential.
The pinion shaft plates, on the back end, are attached to a set of spider gears (like a rear differential). They spin against a main gear anchored to the back of the diff case. This back gear is splined, and that is what the transfer gear is driven by.
As you can imagine, there is a physical lock between the front and rear wheels by way of the spider gears.
Now, it's best to think of what would happen if there were no fluid to provide any LSD function.
The center differential would, then, operate as a completely open differential. This is where things get hairy.
If this were the case, a lack of traction at the front of the vehicle will unload all power to the front. You can't really say torque, because there must be some force at that end. Torque, for this instance, is the frictional force of the tire on the road times the radius of the wheel.
Given a perfect scenario, the fluidless open center differential will distribute power so long as the torque at the wheels is equal (that is, each wheel has equal grip). We'll bar all "what happens when you punch it" because that involves far more than drivetrain dynamics. To quench interest, the fronts will spin first, but due mostly in part to a change in grip that results from body motion. Once the fronts start to spin, it takes a lot less energy to keep them spinning.
Now, add some goo to the diff. The static distribution is still that equal distribution. But as slip occurs (again, dependent on grip at each wheel), the fluid shears and expands. This fluid expansion puts pressure on the plates in the center differential, effectively locking the pinion shaft to the diff casing (and thus the rear wheels).
So when you hit the gas with a completely open differential (like the Cusco Tarmac 5MT diff, or an uncontrolled DCCD diff, or a non-fluid-filled 5MT diff), the fronts will spin due to a change in grip. Like electricity, drive torque (which is different than wheel torque and is defined by Subaru to be the torque applied to the output shaft of the transmission), finds its way down the path of least resistance.
With goo, as this occurs, the plates lock together in an attempt to get the front and rear output speeds to be the same. NOTE, that this is an attempt to get the front and rear output speeds the same, NOT the wheels side-to-side to be the same. So you could completely stop one wheel and double the speed of the other and the front/rear output speed ratio would be the same.
THAT is all the 5MT center diff does. It tries to maintain a static ratio of front/rear output speeds.
By it's very construction, you can see that it would be highly improbable to completely isolate the front wheels from the rear. The only way to do that would be to completely lock the front wheels. If that were the case, power WOULD transfer entirely to the rear, the diff-goo would try to lock the shafts together, but would soon shear and provide no LSD function.
The center diff is, by design, more than just an LSD. It's a NSD (no slip differential).
On a micro-scale, the pinion shaft will start to twist axially before that rotation actually gets to the differential. With a center diff in higher temperatures (as would happen during aggressive driving, or even on a really warm day), this lag in rotation can actually provide enough of a variation in rotational speed to begin plate-lock-up before any substantial slip occurs. By the time the wheels are slipping, the center diff is already damping the faster spinning shaft.
So, to conclude, the center diff isn't "REALLY" 50/50 when it comes to torque distribution. It IS 1:1 when it comes to output speed ratio, or at least it tries to be.
There it is. Glorious in it's simplicity, it's a design that has ruled the AWD world for 20 years. There are only two systems close in efficiency. One is a torsen style helical differential, which again, on a micro scale pre-empts slip through means of physical locking rather than fluid. The other is a clutch-based system which behaves similarly, though requires the use of computers to mandate clutch modulation. Some will say that the computers cannot predict the coefficient of friction, and thus cannot apply the minimum pressure required to do the job of maintaining shaft-speed-ratios, thus having a finite range of lockup (whereas the 5MT fluid based system provides just enough locking to keep the shafts going the same speed). Also, the electronic controlled diffs are reactive, so wheels actually have to slip before lockup occurs (because they reference ABS sensors or some other speed sensor down-the-line from the "diff"). The former is almost indistinguishable to your average driver (though you can generally "tell" when the car is playing with lockup, vs. the 5MT where it's a seamless transition). The latter can be offset by sensory input like throttle position, load, RPM and vehicle speed. If you suddenly go WOT at low RPM and low speed, it'll lock the center diff up the same time it opens the butterfly on the intake manifold, pre-empting a loss of traction.
09 SWP Outback 2.5XT 5MT 263whp, 337 lb-ft