This Class D Chinese amplifier apparently creates a differential (bridged?) pulse-width modulated output with a free-running frequency of about 130 kHz. The rate at which this PWM output changes duty cycle is a function of the frequency of the input signal. The amount of duty cycle variation of the PWM output is a function of the amplitude of the input signal. This particular Class D amplifier does not have any filtering of the 130 kHz PWM output signal, which is required for Class D audio operation. It looks to me like the manufacturer is depending on the inductance and mass of a directly-connected moving-coil speaker to perform the low-pass filtering that will remove the 130 kHz PWM signal, leaving a suitably integrated time-averaged audio signal. Most everyone else uses actual coils and capacitors to perform this filtering.
So, with all this in mind, just HOW exactly do you determine WHEN clipping occurs, WHAT level of input signal is required for clipping, and WHERE do you measure the clipping? Your Golden Ears may give you a clue that something ain't quite right when music sounds okay at low levels and awful when you crank the volume up, but they are subjective, not quantitative, test equipment. It is possible to design a highly efficient (90% or more) Class D amplifier that is also a hi-fidelity amplifier, but I doubt this Asian knock-off falls into the category of hi-fi. The "clipping" you hear is probably normal for this amplifier. Reduce the input signal amplitude until the signal sounds more or less acceptable to your Golden Ears, or consider purchasing a better Class D amplifier. No suggestions from me on that.
The last Class D amplifier I purchased and received (for use in driving a large vacuum e-beam furnace hearth) was sent back to the manufacturer because it was obviously a prototype, but more importantly because it didn't work. The replacement that was sent was less obviously a prototype, but it didn't work either. Bear in mind that this particular Class D amplifier was allegedly designed to work with our e-beam hearth so we would have PID control of the power input and, hence, deposition rate of refractory substances sublimated from the hearth. I guess it was just as hard to design and build a good Class D amplifier in the previous century as it appears to be to do today... although the automobile manufacturers seem to have it nailed.