A great-sounding recording will sound its best only when it's properly mastered to LP, SACD, DVD-Audio, or a high-resolution file. Those formats will reveal the full glory of the music in ways that lower-resolution formats like MP3 or analog cassette always miss. But if you didn't have access to the high-resolution file to compare it with, a great recording will still sound pretty terrific as an AAC, M4A, or 320kbps MP3 file, because the recording's innate quality would shine through. On the other hand, a heavily compressed, processed and crude recording will always sound heavily compressed, processed and crude, regardless of whether it's an MP3, FLAC file, or LP.
Personal preferences are personal -- we like what we like; but there are objective standards for sound quality: low distortion, wide frequency response and stereo separation, and uninhibited dynamic range. If you applied those standards to most contemporary recordings, few would do well. Let's first look at dynamic range compression; once the mix or mastering engineer compresses a singer's whisper-to-a-scream vocal, the whisper will be just as loud as the scream. I love Arcade Fire's music, but their last record, "The Suburbs," sounds like crap. That's too bad; the music is strong, but the sound is not fun to listen to, and an LP or a FLAC can't fix the problems that were there to start with.
No one sets out to make bad-sounding recordings; they all make recordings they hope their intended audience will like. The bands and engineers know that most people will be listening on free earbuds, car audio systems or Bluetooth speakers, so they make recordings that sound good over those things, but if that whisper to a scream vocal was left intact by the mix and mastering engineers you'd hear it in the MP3, FLAC file, or on the CD. Those formats are all capable of reproducing music's full dynamic range, but most of today's commercial music has it soft-to-loud dynamics squashed flat. Apparently, most people like it that way.
If the guitarist was playing a Gretsch Synchromatic 400 Acoustic Archtop, I'd like to hear its unique sound. But if the producer and engineer record the Gretsch through a pickup instead of a microphone, equalized its sound, compressed dynamic range, added digital reverb, and process it to death -- there won't be much left of the Gretsch's sound. It would sound like a generic guitar, which is why I would describe the sound of the recording as "bad." If the engineer boosted the treble to make the sound "cut" better for listeners in noisy environments, it's likely to sound harsh at home in a quiet setting. It would sound "bad" to me. So why not make separate mixes for different formats? Load up the MP3 with compressed dynamics and brighter EQ, and put the uncompressed, less EQ-ed mix on the LP and FLAC releases. Then everybody would get the sound they want.
Most commercial recordings purposely distort and compress the sound of vocals and instruments. And sure, they might even do it in a way that sounds great. That's the idea, after all, but sometimes it's a treat to hear a recording that sounds like the band is in the room with you. Here's a list of some of my favorite recordings.