Knock-in technique for human cells

With this technology, recombination in human cells may be more routine. Wow 15% success rate. How is engineering a nuclease site not increase just random integration? My major complain is that i wish the authors have tried more cell types, particularly normal primary cells. It still seems that the success rate depends on the cell type. Since most cell lines may already have an unstable genome, it may be easier for DNA to be incorporated into the genome. But normal primary cells with minimum culturing should have a more faithful genome, and may be harder to introduce the DNA.