I’ve played around with them before but I did a little research prior to answering this, just to be sure.
It appears that the target hangs freely on its hinged arm and that the sensor is actually located on the surface of the larger supporting arm, measuring the total impulse (I assume) by way of material deflection as the target impacts it and locks into place. This is troublesome for accuracy.
As the target swings upward, the surface becomes increasingly inclined with respect to a linear strike’s direction of motion; at some point, the strike is just sliding past and a substantial amount of energy is lost to back-absorption as you reach maximum extension and have to tense to protect yourself. As a consequence, it seems slightly upward arcing slaps register disproportionately high on this type of machine’s scale.
Furthermore, any time the target isn’t perfectly perpendicular to your strike, a component of the applied force vector is being shunted through the hinged arm rather than accelerating the target.
Basically, it’s not at all a very accurate method of measuring the total kinetic energy of strikes, nor of how concentrated the force is.
It might be a reasonable method of measuring how well one can “snap” their strikes, accelerating near and through the point of impact rather than just building up momentum over a long path (vis-à-vis shorter wavelength of induced pressure waves and increasing inelastic dissipation, a.k.a. damaging soft tissues); however, an accelerometer would function better in this regard.
The machines probably weren’t intentionally designed with these shortcomings: they’re likely the result of reliability, complexity, cost, and “nobody cares” issues with more accurate designs.