As a software analogy, think about a very large and complex, undocumented code-base in a bizarre language you've never come across. Your only experimental tool is to pick a line of code, delete/modify it, and then see what happens. As you learn more, you can do refine your experiments but basically, you're still just 'poking it with a stick'.
We call that a "debugger" around here. You are expected to know the jargon before applying for patents.
What is/are the appropriate scales to study brain function at? Quarks? Molecules? Brain regions? Which ones? All the above? None? We need to know the answer in advance before pronouncing a method "crude", or conversely, "too precise", by some criteria.
We don't know the answer to the scale question but poking "areas" to build up a fuzzy picture of regional function at first approximation is getting us there.
Would the knowledge gained from knock-out/stimulation experiments be more fine-grained if the investigators already had a fine-grained understanding of what they were studying and thus, the cleanest way to study it? Yes.
But breakthroughs are not made where investigators know what they're doing. Breakthroughs are made where nobody knows what they're doing--often crudely, frequently accidentally. The history of scientific advance is messy. Significance of results doesn't map neatly to sophistication of methods/tools.
I wholly agree that neuroscience has a long way to go. But to me that doesn't diminish the achievements of neuroscientists pushing us there, using whatever ethical opportunities available to them.
But obviously it's a lot more than that, and so is electric brain stimulation.