Hey there, here’s my results!
It has been tested on a simplified mesh (16M (~20parts) -> 5M (1part)). I don’t think this changes something (?), just giving the more info that I can. 
i** n blue: parameters that I changed**
in green: best results (between the 2)
in red: worst results (between the 2)
best utilization: 71%
So far the new unwrap seems to work better in “maximal count texture” method and with the “adaptive Texel size” set as default.
But if we try to get more details by decreasing “the minimal Texel size”, the old unwrap method gets better results but not as good as what the new unwrap can come up with. Decreasing “the maximal Texel size” doesn’t get big changes.
Maybe I should have tried with the maximal Texel even more lower to see if legacy unwrap could compete with the new one? I may try it when I get some spare time.
Götz: I just need a UV viewer, that’s all