Atmega4809 RC oscillator not matching 100ppm LMIC frequency

Hi guys,
i have seen a few of you using the Atmega4808/Atmega4809.
For now im using the LMIC_setClockError(MAX_CLOCK_ERROR * 1 / 100); to catch the RX window but i try to find a way to remove this line.

For adding a proper Quartz oscillator i need to add a driver circuit, but this will use more current
Adjusting the CLKCTRL_OSC20MCALIBA, the granularity is too low and does not match the 100ppm s needed(and Jitter)
Use the XOSC32 external crystal as source for the lmic os ticks (#define OSTICKS_PER_SEC 32768 seems to fit neat)

I asked the nice guys at avrfreaks, but before replying there i want to ask you for your experience with this Topic

Why do you want to do this? What is wrong with it?

As will widening the receive window, but a crystal may cost more in the long run as it’s on almost all the time.

That would require some fairly radical replumbing of oslmic as you won’t be able to run all the firmware at that speed so you will end up with two clock sources - one for firmware and one for the RTC(ounter).

FWIW, my 4808 board uses 2 as an error correction.

Thank you, that’s good to know