Most of the applications I have seen with load-cells call for periodically re-taring or zeroing out the reading. Just comparing the reading with a similar load cell at a known load doesn't really tell you that much. You have to un-load THIS load cell and take a reference reading, then load it again, to get a good useful reading from a given load cell.
So in my project, which is for measuring the weight of a 300lb LPG tank, or for the present project, to weigh a beehive, a critical requirement is that the propane tank or beehive is going to want to sit on the load cell for weeks or months at a stretch, and you still want decent readings -- say within 1 part in 100. No one's going to come around and lift off the hive every few hours to zero out the load cell -- it has to remain accurate under constant load, day in, day out.
For this type of application, I have imagined two approaches:
(1) build a balance-lever linkage system that enables a load cell to be presented with 1:1000 of the load. Have the small load spend most of it's time against a stop. Once in a while, when a reading is needed, use a servo or similar mechanical linkage to push the load cell against the load to move it away from the stop. Take your reading, retract the load cell. This way the cell spends most of its time un-loaded, so fatigue and creep don't affect it, and it is always neatly tared to zero before each reading.
(2) build a pneumatic balloon system under the load, like a hot water bottle. Inflate it and seal it up, then just measure the pressure in the balloon. If its hard to keep the balloon inflated all the time, then let the weight rest at the bottom and make a contact closure. When reading is needed, pump air into the balloon until the contact closure opens, take pressure readings for a while, then kill the pump and let the weight settle back against its stop.
I'm not super enamored of either, really, but they're the best solution I can up with for the re-tare problem.