To flesh this out a little and explain more of what I'm looking at I've attached a comparison of the SNR per Tone (the one displayed by the DMT utility) of two ADSL2+ lines. Besides that both are ADSL2+ lines, they are far apart geographically and in performance.
The gaps in both graphs are:
Gap1 474kHz-479kHz (Tone110)
Gap2 548kHz-552kHz (Tone127)
Gap3 811kHz-815kHz (Tone188)
Gap4 824kHz-828kHz (Tone191)
Gap5 1048kHz-1052kHz (Tone243)
Gap6 1255kHz-1259kHz (Tone291)
Gap7 1501kHz-1505kHz (Tone348)
Graham,
These dips, or even nulls, are very familiar - allowing for regional variation. Here's my QLN noise graph:
And here's what a little analysis (confirmed by my wife's trusty portable radio):
Tone, Frequency Khz, AM Station
48, 207, Radio 4 LW,
128, 552, ?
188, 811, Radio Scotland
211, 909, Radio 5 Live
252, 1089, TalkSport
267, 1152, Clyde 2
282, 1215, Absolute Radio
355, 1530, Celtic Music Radio
I suspect yours are something similar.
BTW I have the OR NTE5A faceplate that includes the same REIN filter as both the OR RF03 REIN and the IS-plate, no bell wire, and even a ferrite core on the DSL lead.
Hope this is of some help.