The most variance I have witnessed on this current line is 0.2 dB (9.2-9.4 dB) for downstream attenuation between two different modems - 8800AXL R2 and 8800NL. Though I cannot say anything in regards to Lantiq since I won't let it touch my line as in Broadcom I trust. Don't even talk about MediaTek or I'll scream.
When comparing different chipsets and software, the displayed figure for attenuation may be interpreted differently. But not only that, slight inconsistencies in the modem at the time of manufacturing may lead to some additional (but very minor) attenuation.
I know that on the 8800NL I have, the band D3 seems to be slightly more attenuated by around 2 dB compared to the 8800AXL R2. Though, overall, the difference is not significant enough for it to have any noticeable impact on sync, attainable and the line's overall performance.
I found a nice short but explanatory reply in a different topic:
Line attenuation can only change if the physical charateristics of the line changes, e.g change of gauge of copper wire and it is only meaningful when expressed as a figure at a given frequency. However, you are not actually measuring attenuation, you are extrapolating it from power loss at various frequencies hence upstream and downstream attenuation figures are different. Different routers calculate this in different ways, and a change in tones used or a change of router can be perceived as a change in attenuation.
So, essentially, it is saying that changes you see in the attenuation are down to different modems and down to which tones are allocated. As an example, if your modem suddenly is able to allocate tones at higher frequencies due to a drop in SNR or less interference, the attenuation figure may actually see an increase by as much as a few dB because it is calculated across the tones in use. But if you get an increase in SNR or interference increases by the next time you sync, you may see a lower attenuation value as the allocation of times may occur only at lower frequencies. I believe this is true, but if someone knows I'm wrong, please point this out and correct me.
Which would be the sensible figure? There is no way of really knowing. I would be personally more inclined (biased) to believe the Broadcom figures more than the Lantiq. But that is just me.
I do not want to claim that I know how to estimate the line length on VDSL based on the provided attenuation figures. Though, I would be wary since you will only find out the attenuated distance rather than actual distance due to factors such as different gauges of cable used, and some parts could be copper with some aluminium. But if I wanted to take a guess based on how the line is performing and the approx. attenuation, I'd probably say it's close to 1.8 KM.
Could you upload some graphs from something like HG612 Modem Stats? You should be able to use it with your 8900AX. Just setup the program and use the "current stats" executable to graph all the stats for uploading.