Did you try the g997sang command for getting the SNR per tone? I expect the reason that the some of the data has been stored as averages for 8 tones is because that's how it's required to be sent to the other end when requested (G.993.2 11.4.1 - there's a maximum of 512 groups).
Every time I obtain the data using
g997sang there are quite a few missing groups.
Assuming that means that my connection can't actually use anything those groups, I have added the groups with a value of zero for each of the 8 tones within the group.
The resulting graph isn't at all pretty (see attached).
Now, as when using
"g997dsnrg 1 1", data is reported at the higher frequency groups that wasn't previously reported via my HG612 modem.
Considering my connection can't use these frequencies due to attenuation over 1100m or so of copper from the cabinet, I am starting to think there is some sort of error regarding SNR data (regardless of which command is used to obtain it).
Data for the other graphs looks reasonably similar to graphs from my HG612 (apart from Hlog which now shows a lot more at the higher frequencies than with the HG612).