Welcome, Valery, to the OEM forum.
Unfortunately, almost everything!
We use the 1.024 V internal reference for measuring the voltage from the 0.333 V rms output from the current transformers, and a similar voltage from the emonVs, and this gives a sensible amount of headroom, without losing resolution. Clearly, with the maximum nominal measurable voltage being 1.024 V, but assuming it’s at the bottom end of the tolerance range (“typically” ±4%, with no defined limit), which would be 0.9830 V, it’s not going to be possible to compare and measure a 1.024 V “standard”. You need to be able to guarantee the AVR-DB’s internal reference is well above the nominal 1.024 V. This is why the 1.8 V MCP1501 is used, divided down to 0.900 V using ± 0.1% tolerance resistors.
Note: When designing your cicuit around the MCP1501, do not follow the emonTx4 schematic - the 2.2 µF with 10 Ω in series can cause gross instability and the output becomes unusable. Instead, follow the data sheet examples closely. I believe before long, emonTx4’s from The Shop will have the capacitor removed. See emonVS Power Supply reporting Strange voltages - #27 by Robert.Wall
Hopefully, the next manufacturing batch will implement a better solution.