You cannot objectively analyze the performance of perceptual audio codecs. They exploit subjective perception of audio signals... I haven't dug into your analys

You cannot objectively analyze the performance of perceptual audio codecs. They exploit subjective perception of audio signals... I haven't dug into your analysis but on the surface it appears quite biased against MPEG codecs that exploit masking to determine quantization levels (one of the major advantages being that humans can't perceive quantization error in the presence of loud/noisy content, which can be exploited to use far fewer bits for the same samples!). The closest thing to an objective analysis tool is PEAQ which is also pretty terrible, but its universally terrible so that makes it useful as a benchmark.