To get the basics out of the way - FLAC is, with mathematical certainty, lossless. This has never been in question, aside from people who might argue that 2+2=7. The output of a FLAC decoder will be identical to its input with no variation. You can pretend you're working with WAV files. Or, you can in fact work with WAV files, and you'd get identical audio and spectrograms - just using a lot more disk space. The bottom line is that you can ignore that these files are FLAC since that's irrelevant to any questions regarding the audio itself.
So now we get to the audio itself. The big mistake with your method there is that you are comparing two different tracks. These tracks might have been produced and mastered differently - not to mention they are different pieces of music - so the fact that their spectrograms differ is to be expected, and it indicates nothing.
Many CDs are mastered so that they have sound content all the way up to the Nyquist limit at 22.05 kHz. Many other CDs are mastered with a lowpass filter a few kHz under the limit - this is mostly done to avoid aliasing problems when digitizing the signal.
You are correct in that older music won't have much if any signal above 15 - 18 kHz or so. In those cases you'll see a gradual roll-off on the spectrogram. In newer music, where you have signal up through 22 kHz, but the audio has been lowpassed at 20 kHz for CD, you'll see the "shelf" you described.
Lastly I'd like to say that while looking at spectrograms in this manner can sometimes be a useful tool in determining how a piece of audio was processed, it's really just an academic exercise. The frequency ranges we're looking at have little-to-no bearing on how it sounds to your ears. Take one of the files with sound at 22 kHz, lowpass it at 19 kHz, and you will not hear a difference.
So now we get to the audio itself. The big mistake with your method there is that you are comparing two different tracks. These tracks might have been produced and mastered differently - not to mention they are different pieces of music - so the fact that their spectrograms differ is to be expected, and it indicates nothing.
Many CDs are mastered so that they have sound content all the way up to the Nyquist limit at 22.05 kHz. Many other CDs are mastered with a lowpass filter a few kHz under the limit - this is mostly done to avoid aliasing problems when digitizing the signal.
You are correct in that older music won't have much if any signal above 15 - 18 kHz or so. In those cases you'll see a gradual roll-off on the spectrogram. In newer music, where you have signal up through 22 kHz, but the audio has been lowpassed at 20 kHz for CD, you'll see the "shelf" you described.
Lastly I'd like to say that while looking at spectrograms in this manner can sometimes be a useful tool in determining how a piece of audio was processed, it's really just an academic exercise. The frequency ranges we're looking at have little-to-no bearing on how it sounds to your ears. Take one of the files with sound at 22 kHz, lowpass it at 19 kHz, and you will not hear a difference.