That is interesting. Can you get all 4 e-meters and the gasmeter from the PO output? If so I could implement this in my homey app.
But on the page you linked I can only see S0, actual watt and (aggregated?) E-meter.
Edit: oh wait, is it now possible to get the raw P1 datagram? If so I can definitely implement support for that to make it independent of firmware installed on the Youless device.
Correct, read the part about RAW P1 telegram 
Firmware independent app would be great 
There is also an experimental PO firmware with all e-meters and gasmeter output. Read the thread linked on tweakers.net
Thx! Will definitely check this out. Hope I can make the Homey app firmware independent 
Super news for PVOutput firmware users. Based on the info from @Torch1969 I contacted youless on the raw P1 information in the LS120.
It appears this raw data is indeed available. Downside is that the information is only intended for debugging purposes. This means that the raw P1 data is very slow (it can take up to 20 seconds to get it), that while retrieving the data the LS120 stops doing everything else (no S0, power, gas updates etc), and the data is more often corrupt than complete.
On the upside I got informed by Youless that the new PVOutput firmware 1.4.4 will get the same data interface as the Enelogic firmware. 
That means that if you have a LS120 with PVoutput (-PO) firmware installed, you will now be able to fully use my Homey Youless app.
The firmware is still in beta testing, but if you want to try it out now here is the download link:
http://www.youless.nl/tl_files/downloads/youless-firmware-1.4.4-POb2.exe
I can confirm that it works.
Iāve seen similar slow decrease in graphs. My assumpion was that the Youless firmware is interpolating/extrapolating/averaging to convert water pulse frequency to water flow level. Every pulse represents the same amount of water. Fast pulses mean high water flow, slow pulses mean low water flow. If the firmware does nothing, youād get pulses in the graph instead of a higher or lower graph.
So my guess is that it does something like this:
When the tap is opened, the fist pulse represents the first water intake, but thereās now way of telling if the water flowed fast or slow. It needs the next pulse for that, but that may take some time. It probably assumes slow (delaying the graph a bit) until the next pulse comes and it can get a sense of the speed the water is flowing from the tap.
So, when the tap is fully open for a while, the firmware is seeing high flow and correctly expecting pulses to come in at the same rate. When you suddenly close the tap, the pulses arenāt coming in as expected. So it lowers the expectation for the next pulse, slowly dropping the graph. In the end the total amount will be fine (because it underestimated the start and overestimated at the end).
To put it another way (wish I thought of that first ) It is spreading the pulses over time, based on the frequency of the pulses. At the end the freqency is hard to guess, because you never now if the next pulse will come or not.