Inverter data vs App/Energy Graph

Hi all,
I've noticed a general inconsistency between the values shown on the inverter's meter data tab when compared to the values shown in the energy graph or on the app. Does anyone know which is actually the correct value?
Case in point - according to the inverter's meter data today I have generated 9.1kWh, whereas both the app and the energy graph say I've generated 10.58.
Obviously it doesn't change anything substantive which is the correct value, but I do keep track of overall consumption and having accurate data is useful.
I've queried this with GE before and not had a satisfactory answer yet - there's a known issue that may cover this but it's been up for a while now.
Thanks for any help!

I'm not 100% sure but have noticed differences occasionally.

I wonder if the app has stale data sometimes. Or perhaps the app and the portal use different ways of calculating things (that rings a bell from a while back).

Also I've noticed the energy graph has missing data sometimes. E.g. when you look at the myInverter data on the portal there was sufficient usage in a half hour period for it to draw a bar on the graph for that half hour. But it didn't. Seen this a few times.

I guess if you want accurate stats, measure it yourself and collect the data in home assistant or use GivTCP with powershell or python, perhaps with your own CT clamps and/or raw data from the grid (via your supplier if possible or via a Hildebrand glow IHD I think).

I'm wondering if the difference is some form of dead reckoning. My guess is that the inverter knows how much power has been flowing precisely, and that its the "Generation Today" figure on the meter data page reflects that. That being the case, the inverter data should be pretty accurate as it'll only ever be as far off as it could be between its CT clamp checks.
On the other hand, the energy graph and app might be getting the data every five minutes (needed to plot graphs) but then inferring the usage in between - essentially taking the midpoints of the linear graph formed and using that to calculate usage over the 5 minute interval.
If that is the case, then the inverter data should be good enough. The discrepancy is effectively similar to the way that the "Import today" figure never matches up to what Octopus tell me I've consumed, which is explained by differences in the measurement from the clamps.
At some point I probably ought to play around with GivTCP...