I already answered a related question. I may not know the most definite answer, But I read that iTunes Match uses Gracenote's MusicID program. It seems MusicID uses the following techniques to identify songs.
MusicID-featured Identification Technologies:
CD Identification: Using the CD’s unique Table of Content (TOC), MusicID can identify CDs released around the world. Even with multiple pressings and release dates, patented fuzzy matching enables accurate recognition.
Stream Identification: Using an audio waveform fingerprint technology, MusicID can identify music with short audio samples anywhere in the song.
Digital File Identification: MusicID can identify individual songs or albums in a user’s collection regardless of source, format, or tag information. Utilizing a multi-step process which combines stream identification with textual information analysis from the file’s embedded tags, file name, directory path, and other related files in the user’s collection, individual or groups of songs can be matched to albums.
More here.
To answer your question, they perform some fuzzy text matching algorithm to map meta data and also perform audio waveform analysis.
If you notice in your screenshot, the files that were uploaded were ripped at 128kbps. This causes enough audio artifacts in the file to throw the acoustic fingerprinting that was used when iTunes Match first launched. It has since been improved and issues like this with lower bit rate files are now much less common.
Best Answer
I solved this by creating another iTunes library and switch between the two libraries: iTunes Match and Library on NAS.
See: https://support.apple.com/en-us/HT201596