Almost all cartridges and printers (besides some Canons and HPs), use an open-loop ink tracking system. That is, the printer keeps track of how much ink it ejects from the cartridge, and deducts that from the chip's internal counter. There is no system to actually measure the actual ink in the cartridge on refillable carts, and none on most OEM carts. Since the ink wicked out of the head without the printer knowing, it did not update the chip, and thus your reading was wrong. As was stated elsewhere, refillable carts are much more prone to wicking than OEM, so even if your printer works fine with Epson carts, you may have problems related to dirty systems with refillables. It's unfortunate but true.
(Some Epson carts have an air sensor inside that detects when ink is depleted and will lock the cartridge at that point. Canon uses an optical crystal to see if there is ink in a liquid reservoir, but then goes into open-loop for the remaining ink in the sponge. On liquid-filled HP carts, the printer detects when the pump can no longer pull any ink out, but is open-loop until then.)
Oops...wrote the above without throughly reading your above post. Sorry for repeating some stuff.
only estimate what is used. From what, exactly? What kind of algorithm is used to estimate ink usage? How are the Cone cartridge chips different from the OEM cartridges, which seem to do a pretty consistent job of monitoring the ink, in my experience?
From my experience with the R2880 refillables, they are quite accurate when printing without problems. It's only if you didn't completely fill the cart, or have leaks, that you run into problems. I assume both OEM and refills use a similar starting counter value, which is then modified by the printer whenever it knowingly uses ink (it knows how many picoliters each drop ejected is, and it knows approximately how much ink is used in a cleaning cycle). The cart itself is pretty dumb; only storing a value which is modified by the printer.