Close lines - bad results

[ghuotvez]

I am using “Close lines” tool for some line landforms like lateral meltwater channels.
I used 2000m for the “Minimum tolerable length” and 3000m for the “Minimum tolerable proximity”.
But the features which are kept as 1 are much closer than 3000 m.
Strangely when I changed the “Minimum tolerable proximity” to 1000m, less lines were preserved than when I used 3000m.
Is there a limitation for the Minimum tolerable proximity? Should it be less than Minimum tolerable length?

[ghuotvez]

Tested it with Abitibi data and I do get weird results. Tested it with Rebord Ravinement, while the length is being treated well the proximity one doesn't get right.