Your pipeline seems to be doing reasonable things, though I'm not really sure on the biology so I can't say for sure.
I think if you're not getting the results you expect you need to
1) Make sure you agree with the probability maps you're getting from Ilastik
2) Make sure you agree with the CellProfiler segmentations of your cells (is it finding the right cells, are some getting missed, etc)- I often find adding an OverlayOutlines+SaveImages modules is nice at this stage so you can quickly flip through the segmentations.
3) Think about whether or not mean intensity is the right metric to judge viability- again, without knowing the background of the experiment or even what dye you're using it's hard for me to say.
You could also think about using CellProfilerAnalyst to look at your data; you could add the dose information in the Metadata module, export the measurements to a database and then use the graphing tools, plate viewer, etc to see if you can find something in the CP measurements that nicely correlates with the dose-response that you expect.