Resazurin assay inconsistency - inconsistent readings over the plate (Jan/11/2011 )

I'm trying to perform a Resazurin assay to assess the change in growing rate of cancer cells with and without treatment.
The problem is that when I add a similar number of cells in all wells (96 well plate), and the day after add Resazurin (without any treatment to any of the wells), I get different readings over the plate: The two rows which are closest to the top of the plate have the highest readings and after that the lanes which are on the side of the plate.
The difference between the highest and lowest reading are of 40%. (100% is top row 60% middle of the plate)

I thought that the CO2 is more available to these wells, so I used an adhesive strip instead of the plastic cover on the plate, but that didn't change anything.

Here is the protocol I use:
1- For attached cells - detach the cells with Trypsine, Dilute with DMEM 10% FBS and spin down (I usually use BxPC3 and Panc-1 cells)
2- Discard supernatant, dilute in DMEM 10% FBS, count cells, divide cells equally to a 96 wells culture black plate in 200ul DMEM in 10% FBS (with antibiotics). (I tried using anywhere between 5000-30,000 cells/well)
3- Incubate the cells (37C, 5% C02) for 24hrs, treat the cells with according reagents and wait 24-72hrs
4- Add 20ul Resazurin (R&D systems) to each well, wait 2-5 hours and read fluorescence at 590-530nm.

I will gladly receive any ideas and suggestions on what I'm doing wrong,

Joel

I would avoid plating cells in the outer wells of the plate since they are prone to evaporation and other effects. I usually plate the inner 60 wells of a 96-well plate. Try running a plate with a fluorescent dye in all wells (no cells) to see if there is variability among the wells. Better yet, get a QC plate and see if there is the same variability exists.