In terms of efficiency, calculated as ( speed_tester / sync ), there seems to be a 3% improvement in downstream efficiency, exactly as predicted, and a 5% improvement upstream which is a bit surprising, although these are rather rough figures. These were based on the highest speed tester values only, taken from speedof.me, not on the mean. Although speedof.me does exaggerate, I feel sure, I am comparing like with like and the comparison is based on the ratio of efficiency values not on absolute differences.
So every little helps, it works for BTW 21CN ADSL2, and if it works for you the definitely go for it. It should always be the first choice. Not applicable of course if you are on VDSL2 as there's no cursed ATM, no PPPoEoA and therefore no inefficiency to worry about so nothing to improve.