Torque is what matters.
If you test bending in the middle, you put 70 pounds force (approximately 300 newtons) in the middle, and support the ends firmly. That means that if the weakest point is in the middle, the torque on that point is (2 x 150 newtons x 80 mm), as the length of the lever arm to the support at the edge is about 80 mm. We'll ignore the fact that the test support is really a little bit inside that, and assume that the subject is supported right at the edge. Note also that the force is 150 newtons, which is half of the 70 pounds force used to break the phone, because the force is opposed evenly by two supports. Their equal force is then summed, which is why our total torque has that "2" scalar, giving us a total of 24 newton-meters of torque.
If we bend off-center, such as half-way towards one of the ends, the forces on the test supports are no longer equal. Our lever arms are now 120 and 40 mm, and the force would be unevenly distributed as well. The force is distributed inversely to the length of the lever arms, so the short arm, being 25% of the length, now supports 75% of the load, which is 225 newtons. The long arm supports 25%, which is 75 newtons. This gives us a total torque of (225 newtons * 40mm + 75 newtons * 120mm), for a total of only 18 newton-meters of torque.
Since testing off-center actually applies less torque to the test subject, the question then becomes one of whether the weak point is really 25% weaker than the rest of the beam.
However, we can also compute the torque on the supposed weak point during the center test. In that case, the lever arms can be computed as though they behave as a typical lever, scaling the force. they apply. The longer lever would be a class 3 lever, which would reduce the effective force of the test to 100 newtons. On the other hand, the shorter arm would behave as a class 2 lever, increasing the force to 300 newtons. The total torque on the weak point during a center test, then, is (100 newtons * 120mm + 300 newtons * 40mm), which is again 24 newton-meters.
If the weak point were really weaker than anywhere else in the phone, it would break during the center-loaded test. Looking at the pictures from Consumer Reports, though, that's exactly what happened. On both the iPhone 6 and iPhone 6 Plus, the most significant damage is at the edge of the volume buttons closest to the center.
However, it's worth noting that the Consumer Reports test was conducted until the screen detached, even if that happened after the phone itself was permanently deformed. Looking at other pictures of bent phones, their screens have not separated from the cases, so they likely used less force to deform. Bending to separation, though, provides a consistent point of comparison to other phones, which may have internal damage even if their cases return to normal.
Disclaimer: I am not a physicist, and not a test engineer. If my math or methodology is incorrect, please feel free to tell me why.