The 'usual' master socket ring wire tricks are the usual ones because they are pretty much all you can do to increase the stability of the connection! I take it there's nothing new on
http://www.kitz.co.uk/adsl/lowSNR.htm that you haven't tried already?
In fact, until I read what you'd written below the graph I was just thinking how good your graph was looking!!
I'm presuming from reading perhaps too much into that graph, that you aren't having any stability issues at present? I'd be interested to see the line stats too, specifically your error counts.
The sharp drop at 270-ish is the 300kHz "pilot tone", which is present on all ADSL1 connections. The other significant drop is around 400kHz, I don't know what that could be caused by - perhaps tuning a long wave radio into ~400kHz and trying to trace the source of that interference might help, but I don't know. And on the SNR graph it doesn't drop at the same place, so perhaps a red herring of sorts! There certainly doesn't look like there's anything to worry about on that graph, or indeed much that can be done to improve matters at all!
The overall shape of the graph is very good, and shows a nice clean line.
Due to the nature of the beast, it is totally impractical to expect to be able to get rid of all the imperfections and interference on your line - the SNR per tone graph looks like it bodes well for a reasonable increase in speed when you go to ADSL2+, since the high frequencies don't appear to be tailing off too much.
Your line is in much better condition than mine is, and I'm syncing at 9.6Mb on Be. Perhaps someone else will come up with something I've not thought of