This is quite an interesting article. I'll offer up a few comments.
First, the recordings are all quite old and were mixed and originally mastered with vinyl in mind. The CD mastering was "an afterthought". There have been previous comments about how "poor CD was when it first came out", and some of the responsibility for this lies with lazy (re)mastering for CD.
If someone has the CD's to hand, it would interesting to know whether they are AAD, ADD, DDD recordings. Analog master tape tops out at around 80dB of dynamic range, and those recordings aren't the best thing to showcase CD with..
Using a soundcard equally disadvantages both vinyl and analog outputs from digital devices. The difference between the EAC vs. CD results (1st table) suggests something isn't quite right - the Sony SCD machine is quite capable and this suggest that this soundcard is introducing significant distortion... [FWIW: If I were doing this experiment, I would have stopped at this point and gone looking for a better sound card - or a proper bit of
calibrated test equipment.]
It is too bad that middle C is below 500Hz. The comment about "the majority of the spectrum" is quite amusing given it's close proximity to the word "misleading" when considering both the power spectrum and frequency content of music/human voice.
"relative dynamics" isn't a technology issue as it's really set "in the mix" (the difference of 0.5dB isn't worth shouting about..)
The "waveforms" are interesting, but what's more interesting still is this leap of imagination:
This finding supports my own subjective impressions comparing the CD against the LP. I much prefer listening to the LP over the CD on my system. It's a fairly classic example of "cherry picking" the data to support a conclusion.
"Mick's Blessings" - the original vs remastered is actually a sad reflection on the "professionals" rather than the digital tools. Believe it or not, this "gain riding" madness happened with vinyl too when people were trying to make their single sound the loudest on the jukebox.
Overall, the most significant blunder in this article is to "join the dots" between the statistics of the data and to extrapolate that to the "explanation" for the authors subjective opinions. Moreover, there's a "sample size of one" for the subjective data. It's quite a poor science experiment and would never pass peer-review - it's good enough to publish in Stereophile though.