This is a legitimate concern.
If the high-frequency PWM is driving to the voltage rails, then it should be able to generate voltage swings as high as an audio band PWM.
I realize you can’t actually fit a speaker in the Arduboy, but it would be informative to try one of the micro 8-ohm speakers with the various types of sound libraries. If the variation goes away while driving a speaker, that would suggest the difference is due to some feature of your piezo sounder.
Finally, one key feature of a high-frequency PWM module is the ease of making volume changes. It would be wise to be certain that the library that seems too soft doesn’t have a volume setting buried somewhere that has been set to less that unity (full volume).
If you have a detailed level of control, try setting up an audio band square wave PWM at several useful audio frequencies, scattered between 200 HZ and 5000 Hz - the main hearing band. Then use the high-frequency PWM library to produce square waves at the same frequencies. Electrically, the results should be identical. Thus, so should the sound. If they are still different, it’s time to break out the oscilloscope and check output drive levels.
If the square wave sound is the same, but other high-frequency samples are different, then you may be facing a difference problem - namely, that sound loudness is usually proportional to RMS power. Square waves (and audio PWM source) have about 40% more RMS power than an equivalent frequency sine wave (hi-freq PWM sources) , and as a result, will always sound louder. (They will also sound different.)
If that’s what you are encountering, then you have a tough decision to make, because no one solution will give you a consistent audio level between the two types of sources. If the audio PWM is fine, then hi-freq PWM will be lower. If hi-freq PWM is fine, then audio PWM risks being too loud.
I realize that the hi-freq PWM will generate nicer sound - it’s engineered to do that! But hi-freq PWM has to usually use pre-calculated sine wave tables. Those typically drive up the size of the library. That’s a key reason for people looking at this to know how much memory that audio driver takes from program memory and from RAM. Some programs may happily devote that space to the sound library, while other programs might need every last byte of program space they can scrounge for their code and screen data.
If you are going to do a survey, at least include all the relevant information so that any decisions are made while aware of the possible trade-offs.