Michel,
There is a chip errata published on Sharp's website viewable here:
http://www.sharpsma.com/s...H7A400-SMA03049A.pdf
If you look in section "Real Time Clock" you will see a subsection entitled:
"13. RTC accuracy may be affected by polling for the RTC value"
This is the cause of the wandering clock. We have observed this behavior at Logic as well and are awaiting a silicon fix from Sharp.
If you need a more immeadiate solution, we have _lots_ of experience with other Real Time Clock (RTC) chips. You can work with Hans and his team to add a cheap but accurate external RTC.
Best regards,
--mikee