Hello yet again Floor,
Bearing in mind that the Samknows / Kitz calculations can only extract one Grid Ref position per postcode,
you might just as well do an alpha-sort on postcode and then take one random number to use in the calculations.
The only time that method would fail is if you positively know you're working across exchange boundaries as then Samknows will pick up the different post code co-ordinates for the different exchanges.
Without trying to impute BT's internal databases if they are available (as opposed to the public line performance database) there are so many variables, such as inter-PCP and PCP to SCP tie cables, and chances for human error that I really don't think it's worth even contemplating such an exercise. In any event, once you start looking at individual twisted pairs and possibly crossed pairs, with or without cross-talk problems, with or without corroded aluminium joints, in a cable run, it's almost inevitable that quite different performances could be achieved on each pair.
All the above is one reason BT are usually quite non-committal as to the actual performance that will be achieved. Having said that, if you keep persevering and are as determined as I have had to be out the sticks, you can often find that BT can swap D side and E side cables to improve a line to the local average.
If you are looking to improve lines in an area perhaps using my method of plugging in my own modem to gather existing performance data in a few properties would be a more suitable solution.
Kind regards,
Walter