When I asked HKJ to put this battery on his review list, I had only a batch with fairly consistent capacity ones.
A bit later, I had more pieces of them (40 in total), and as I learnt, it is kind of a roulette.
If anyone interested, I can show here mAh and mOhm readings for them (mOhm readings are probably less accurate), which I got with a Miboxer C2-4000 charger (discharge current was automatically determined by the charger, usually around 0.55A). Most discharged mAh readings were in the range of 960-1015 acquired by this method.
Out of 40, I got only 2 pieces of the ‘dud’ cells, which were ~770mAh and ~820mAh respectively.
Usually, if this kind of battery already reads below 1.15V on arrival, it is not a good sign (guessing from the graph, the lower capacity one was exactly like this - this kind of ‘duds’ has much more IR at low charge state, which is confirmed by the much-quicker-than-normal voltage rise compared to the other cell). Good cells are almost always above 1.15V on arrival.
HKJ was ‘lucky enough’ to pick one dud from far less samples than me.
As I look at the comparator, variances are also present in IR, not just capacity, as the lower capacity cell seemed to handle high currents a bit better before reaching end of discharge.
Keep in mind that they aren't manufactured with high quality control. I wouldnt call them excellent with 2 such different capacities..
The max capacity of the top one looks definitely interesting, but that is off only 1 test.. Therefore dangerous to recommend them at this point without more tests, seeing already this kind of difference between just 2 cells.
I need to admit, that the manufacturer has quality control issues. I checked through my sets, and experienced that 'low self discharge' ability is variable. Most of them could hold charge over 4 months, but some of them was already below 1.2V, and I found even a few ones, which were in deeply discharged state (e.g. 0.95V).
The recently reviewed PKCell 1000mAh seems to be almost the same (except the end of the discharge curve, which is a lottery, it seems).