I use MTT which is very similar to the XTT assay except for the solubilization step in the end. The variations/standard deviations have often been the problem with this method and I am so far not able to reduce it as much as I would like! However, there are few things you could possibly watch out for. Firstly, the cell number when you seed.. This does make a difference! Also, make sure you have not more than 200µl per well in a 96 well plate. If you do, then adjust your final XTT concentration accordingly. shake the plate well on a rocker for about 10 minutes immediately after adding the XTT and also before reading the plate.
Yes, to mix well after adding the reagent is crucial! To control the cell number after seeding is quite difficult. But to reduce cell number variations in your duplicates, its better to have the cells in high to medium dilution (in less volume of media) before you seed. I cannot think of a better way. And you can of course use multi-channel pipettes if you are not using them already!
And the incubation time is certainly different from cell to cell but a solid 2 hours incubation is good enough for any kind of cell line I think. I have worked with 3 or 4 different cell lines both normal and neoplastic.