Jezz,
Sounds to me like drewcrew6 is right on it. You either have a calibration issue, a load cell size issue, or a design issue.
First, if you are not already checking the calibration of your load cells, this is where you need to start. When you do your calibration the more known weight that can be applied the better, so apply as much as you can. One technique in your situation might be to fill the tank with a known quantity of clean water and calculate the weight using 8.333 pounds per gallon (or 2.2014 pounds per liter).
Second, check that your system is right for the application. A properly sized load cell system will take into account the dead weight of the empty silo itself and then the full weight based on your heaviest/most dense product. If you know these two factors, any reputable load cell manufacture can quickly tell you what size the load cells should be. Someone had to specify (or at least agree to) the acceptable resolution for your system and it is possible that +/- 0.5% accuracy was deemed acceptable at that time. Anyway, you need to find out what the specs on the load cells are and see what the accuracy and repeatability are.
Last (and least likely), check your system design. There should be no external loads on your silos. Nothing leaning on them, no wind allowed to hit them, and no major temperature fluctuation (although many load cells are temperature compensated, I don't know about yours). Are there multiple load cells for each silo? If so, has each been tested independently? If not, your system probably pivots somewhere and you are really only measuring half the load. Check for binding.
Steve