I talked with a friend that works in Gang Bao Lab at Rice. She is working on the off-target analysis of CRISPR/Cas9. She told me most of the problems from CRISPR seen in experimental labs seem to fall in two categories, efficiency and specificity/accuracy.
The efficiency of CRISPR is highly dependent on the sgRNA sequence and some factors can be explained by the formation of RNA hairpins or other low energy structures that may impede the binding to the complex. It may also be caused by the packing of the DNA and the chromosomes, unwinding, methylation. Some of this features can depend on the cell type.
The specificity problems occur due to binding to homologous sites, but it is difficult to predict which homologous sites. CRISPR-Cas9 may cut non-homologous sites but it may be difficult to know given present sequencing technology.