The state of the wiring will certainly have an impact on the attenuation, but the most significant impact is simply the line length. In the example above, the OP went from 8dB down to 7.5dB - a pretty negligible difference, but the sync speed increased threefold.

I don't know exactly how the attenuation is specifically measured for ADSL connections, but suspect it is a figure obtained from a low frequency (close to voice or DC) signal. As such, it isn't affected much by line stubs and other devices putting noise on the line.

When the actual ADSL signal then gets put on the same line, it's far higher frequencies are influenced much more by noise / reflections etc, with the final result being the sync speed scaled back in order to achieve a stable connection.

Doing an isolation test (exactly what you are doing) is the perfect start, because it's eliminating lots of variables straight away.