The curve for the 17HS19-2004S looks sensible to me. Selecting that motor in the EMF calculator at https://reprapfirmware.org/ and selecting half stepping, the calculator predicts that the torque will start to drop off at 2.9kHz (low slip angle) or 4kHz (high slip angle). The figure for high slip angle can be compared to the pull-out torque curve, and it agrees reasonably well.
The curve for the 23HS30-2804S looks completely wrong to me, as if it has been measured at a constant voltage of 3.2V, not at a constant current of 2.8A with a driver supply of 24V. It doesn't show a flat area at all. The calculator predicts that loss of torque will start at 1.2kHz (low slip angle) or 1.6kHz (high slip angle). Whereas the curve suggests that torque falls right from the start of the curve at 0.5kHz.
I also looked up the 23HS22-2804S which is a shorter Nema 23 motor that I use to test Duets at maximum current. The calculator suggests that torque should start to drop off at 2.2kHz or 3.1kHz. This fits reasonable well with the curve published at https://www.omc-stepperonline.com/download/23HS22-2804S_Torque_Curve.pdf.
