This doesn't help identify the cause, but it gives you more comparison data...
There's an interesting user that you might like to compare yourself to over on this thread:
http://forum.kitz.co.uk/index.php/topic,16512.0.html ("G.INP in full flow").
There, you can see a line that becomes afflicted by FECs from around 10pm through to 3pm, day-in, day-out - but only for the last 10 days or so.
Whatever is causing those FEC's can be tracked further into the other FTTC error-recovery mechanisms: particularly by looking at the G-INP graphs titled "G-retransmit tx", "G-retransmit corr", "G-retransmit UnCorr". Those graphs give an indication of where G.INP is doing its job, and re-transmission is coming into play (retransmission, of course, is another form of "CRC that didn't happen" ... and the "UnCorr"s are the retransmission failures that *do* become CRCs).
Ultimately, the G-LEFTRS graph is the one that identifies when the retransmission mechanism is being called into action too much, with a detrimental impact to throughput. That maxes out at 60 per minute - which "tenbyboy2"'s line does a lot. Your line gets an occasional "1" per minute, so isn't too bad.