The Americans have a long standing fascination with measuring 'wheel horsepower', which accounts for the low figures. Problem is that measuring horsepower at the wheels doesn't account for many other variables that will affect that reading. Tyre pressure and temperature will alter the reading, the position of the car on the rollers will alter the reading, the tension of the straps will alter the reading. Even throwing a couple of bags of cement over the driven wheels will increase apparent power.
It's like trying to reliably measure distances with a ruler made of elastic bands.
In the uk, dyno operators try to reduce the potential errors by measuring the rate of deceleration of the rollers on the run down after each run. This gets a rough measure of the effect of all these variables, and factors it into the final readings.
That car was taken off the rollers multiple times, so whilst there are increases, those numbers should not be treated in any way as 'definitive' since you have no data on tyre pressure, roller position, ambient air temp and pressure, strap tension or any other variable capable of changing the readings.
Removing the Haldex fuse was something I had to do in my old Mazda 6 if it got a puncture. Yes it does disconnect the rear driveline, but I wouldn't want to stake my life on it staying disconnected. Given that distance between the clutch plates is absolutely tiny and even the slightest bit of contamination in the oil would cause the Haldex unit to drag a little and perhaps even lock entirely, transferring torque to the wheels, I'm not sure I'd be risking life and limb and testing the car in that manner. Not only does it not give representative data, (no-one ever drives these cars with the rear driveline disconnected), but I'd certainly be wary of the possibility of getting 1.6 tonnes of Teutonic hatchback in the face if the rear driveline does lock.
The products clearly work, but take the absolute readings with a pinch of salt.