The "cut-and-paste" analogy is perfect here. Imagine you discover super-complex code for which you have only a rudimentary understanding. Now imagine replacing sections of code from one program with sections from another program and calling it an improvement before even testing the code!

Older versions of genetic engineering are like pasting code into a random section of a program. CRISPER-CAS9 is more like replacing one section of code with another. Better? Sure. But we still dont really understand the system.