Ah, yes. It was your circuit that was clearly showing an underlying sinusoid, towards the higher frequencies, in the Hlog QLN plot. Unfortunately I am unable to think of anything that would cause such an effect.
Hmm . . . I just typed that last sentence when I had a sudden thought. A number of years ago I performed a series of experiments which consisted of harvesting the data and creating the four snapshot plots (Bit Loading, SNR, QLN & Hlog v sub-carrier) from a circuit whose length (between VTU-C and VTU_R) was increased in 100 metre lengths, up to a maximum of 600 metres. I was using two reels of three pair, CW1308 specification, cable of length 100 metres. To achieve any length greater than 200 metres I was using more than one pair in any one cable. For the full 600 metres, all three pairs were in use . . . VTU-C, reel one blue pair, reel two blue pair, reel two orange pair, reel one orange pair, reel one green pair, reel two green pair, VTU-R. As the total circuit length was increased, so was the coupling between the pairs in the cables. A new term, SIXT (self-induced cross-talk), was proposed for the effect seen.
And now that I have typed that paragraph, I realise that the effect was similar to that seen with a bridged-tap and not an underlying sinusoid, as seen with your circuit.
[Edited to correct the punctuation and an incorrect statement.]