My gut says that determining SOC by coulombs out vs coulombs in is most likely the best way of doing things. It is likely the least inaccurate method. In order to improve those imprecision, we need to understand their sources.
While charging the "coulomb counter" needs to know what percentage of the electrons going into the battery are actually charging the battery or are being spent on heat and gassing. The more accurate we can estimate that "charge efficiency", the more accurate our SOC readings will be.
Looking at the 1st graph below, we can see that a charge from 50% SOC to 100% SOC would likely be approximately 75% efficient. (area under the curve). If, however, a charge from 75% SOC to 100% SOC would likely be more like 50% efficient.
To make things a little more complicated peek at the 2nd graph. If you are charging at lower current levels, the curve shifts to the right due to lower I2R losses. Now the 50%-100% number is more like 80% efficient and the 75%-100% is like 70%.
Not confused yet? Look at the 3rd graph.of discharge efficiency's.
It would seem to me that keeping track of every electron would require pedaflops of computational power and is just not worth the effort.
Let us be happy with an approximation of SOC and realize that is what it is.
On a monthly basis I record KWH in and KWH out of my battery bank. Over the last 2 years it has consistently been 70% +/- 5%. Kinda makes sense because I routinely only discharge to 80% SOC and charge at about 8amps.
It would also help if MidNite had a more accurate 100% SOC re-calibration point. (see separate post on "float current charging")