It’s my understanding that the purpose of upsampling is not to create new and/or better data, but to move the digital filtering to a much higher frequency so that the steep slope of the filter does not affect the audio band upon conversion to analog.
Upsampling is normally combined with digital filtering to make the analog reconstruction filter easier to build consistently by reducing its order (complexity). Often the digital filtering is linear phase and can correct for the phase response of the reconstruction filter giving essentially perfect phase response overall.