Hero, that circuit is VERY similar to what I tried. However, as the battery voltage started to dip, the cap was slowly discharged. Like I said, I tried to use a diode on the cap input (between the 100k ohm resistor and the common point in your schematic), but even the leakage of the diode was enough to discharge the cap. Not only that, but the comparator's (LM339) input impedance was too low (even though it's in the M ohms) and would also discharge the cap slowly.
I even tried to buffer the cap by using a TL082 op-amp (gain of 1) since its input impedance is 10^12, but in the end, the diode's reverse leakage was too high. Because I'm charging at such a low current (0.3A), it takes a while for the delta peak to show up, and that time is long enough for the cap to discharge.
I even tried a 1000uF to lengthen the discharge time constant, but it wasn't enough. If I used a diode with extremely low leakage current, I would still need something like a 10,000uF or higher capacitor to make the discharge time long enough to be negligible. Don't forget, I am trying to sense delta peaks in the 2-4mv range over a 5-10 minute period so
any current flow has to be taken into consideration. Input impedances and tiny leakage currents become a big factor. If I had "ideal" components; diodes with no leakage at all, and comparators with infinite input Z, zero output Z, and rail-to-rail outputs, then this circuit would be a piece of cake.