Maybe another member can enlighten me about that graph.
Not sure why he uses force instead of torque but it doesn't really matter for the point he is making. The 2 force vs angle graphs are a bit different than the standard ones you generally see in the literature but the bottom graph is force vs current vector phase angle with respect to the d-axis for an IPM (most graphs I've seen are with respect to the q-axis).
For a surface mounted rotor, there is zero reluctance torque so all you have is the "lower frequency" grey line. This tells you that to get maximum torque out of a surface mounted PMSM, you need to have your current vector on your q-axis (90 electrical degrees from your d-axis).
For an IPM, you now have reluctance torque added in (the second "higher frequency" grey line). Reluctance torque only depends on the phase current and the change in inductance with respect to rotor angle. Because of this, the reluctance torque in IPM's is always twice the electrical frequency of the motor. So if you had an IPM with no magnets in it you could still generate reluctance torque if you changed your current vector phase angle to 45 degrees from your d-axis.
So for an IPM when you add the torque from the magnets and the torque from reluctance your maximum torque occurs when your current vector angle is somewhere between 45 and 90 degrees from the d-axis (depending on the design of the rotor and the load).
The problem with this is that those 2 peaks don't occur at the same current vector angle. Their new motor design with slip rings allows them to align the PM and reluctance torque peaks so they occur at the same angle.