Tough question... it will depend on your phenotype. Short of reconstructing the proper control, you can use your wt cells. Normally for these kinds of experiments, I would have started with a single clone that best reflects the behavior of your wild type population, as the selection of single clones incurs a large founder effect risk. That being said, alternatively, if you have selected multiple clones for your knockout, you can use three independent clones and compare it to your wild type cells to minimize founder affects.

Ultimately, it will depend on the experiment and results...but given your situation now, I would try the above ideas....

I agree with Nicolas for his suggestions which are time saving for you! In future, you can try creating controls (+ve and -ve) based on assays you are performing to have clearer interpretation of the results. Lets say knocking out another gene which will give expected phenotype (+) and knocking out another/unrelated gene without any phenotype (-) using the same approach. All the best!

I think integrating the same vector with Neo and eGFP in a random genomic location would be your closest control