Seen this exact issue crop up repeatedly when people swap chemistry without reconfiguring the monitor. The BMV-712 (or whatever you're running) will have been calibrated for lead-acid charge/discharge curves — lithium has a completely flat voltage plateau through most of its SoC range, so voltage-based estimation is essentially useless. You must be running coulomb counting, and even then the Peukert exponent needs resetting.
Specific things to check:
- Peukert exponent — set to 1.05 for LiFePO4, not the 1.25-ish you'd use for AGM
- Charged voltage threshold — needs to match your BMS termination voltage exactly
- Tail current setting — lithium absorbs very little current at full charge, so this wants to be around 1-2% of capacity
- Synchronisation — have you manually synced it to 100% after a full charge since the swap?
My shepherd's hut setup had similar nonsense readings when I first fitted a Fogstar Drift 100Ah — the monitor was convinced it was perpetually at 73% regardless of what I did. Turned out the charged voltage threshold was set 0.3V below where the BMS was actually terminating, so it never registered a full charge and never reset the coulomb count.
Also worth checking whether your shunt is properly sized for lithium's higher potential discharge rates. Lead-acid users often run undersized shunts because the batteries self-limit. Lithium won't.
What monitor are you running and what BMS is in the new battery? That'll determine exactly which parameters need touching.