A capture of the QLN from DSLsats wouldn't kill you!
And it would certainly give us a better view of the total spectrum.
Though, from how you're describing it, it seems to be OK.
However, there are a couple of distinct areas around the 200-300 and 3100-3200 tones where the clump of values have a general base line value but can peak many times over the range of tone numbers (e.g. 10-15dBm/Hz variation) and then back to normal outside these tone ranges?
I believe this is normal for those lower tones. You have to remember that ADSL also will also almost certainly exist alongside VDSL (FTTC) deployments, and ADSL2+ has 512 sub-channels (up to a frequency of 2.2 MHz). The peaks could be where Power Spectral Density (PSD) comes into play as it is less than that of ADSL (so ADSL can actually impact VDSL as opposed to the other way around). For myself, I see peaks from between 200 until around 500 where the line evens out.
I could be COMPLETELY wrong, but I await someone of experience to confirm this.
Edit:
I should add that the above doesn't confirm or explain why tones 3100-3200 (~13.37-13.80 Mhz) are impacted. Not sure which equipment would operate on here.