I think we're rather in the area of guesswork. I follow your reasoning, but I think I prefer to leave the uncertain areas blue, rather than (possibly) guessing wrongly.
I'd agree with that for the bit-loading graph as that is truly dynamic & picks up changes such as bit swapping etc. during the life of a connection, as does the SNR graph.
I suppose that the real question is - what use is made of these graphs, and how important is it to distinguish between up and down in the uncertain areas? I confess that I'm doing this in a rather mechanistic way, without really understanding the value of the information. If you consider that a best guess is better than no guess at all, I'll happily go along with that.
Hlog & QLN are very important as they basically dictate sync speeds etc. as they do reflect actual physical conditions at the time a connection syncs.
Any significant changes in these (attenuation & noise) have quite a dramatic effect upon a connection's speed/stability.
e.g. I often saw dips/valleys in my Hlog graph when my connection was regularly "playing up". The graphs were partially the proof of the problem that prompted so many engineer/technician visits without being charged for any of them.
I haven't seen a single sudden dip/valley in my Hlog graph since my connection was finally repaired.
How important it is to distinguish between DS & US for the "shared" tones as seen on VDSL2 connections & not seen for ADSL connections is debatable, especially as Huawei DSLAMs only report DS QLN & Hlog data anyway.
I chose to make a reasoned sort of stab at it for ECI DSLAMs.
Please don't let this particular matter detract from the excellent work that you have carried out so far & will hopefully continue to carry out as you further develop your program.
TBH, it is probably way down the list of priorities.
I only mention such things as I am aware that you don't have a VDSL2 connection of your own to play with, as my attempt to assist with your overall perception of VDSL2 matters in general.