I do not think anything can be truly accurate or perfect online, the nature of bots, broken links, and again too many variables. But it is 100% improvement on the old search console. In short it is unquestionably more accurate than the old search console. The old search console had too many variables. The new search console, gives us more succinct information.
And on the cross comparison of data between old search console and new search console - I think this would have negligible utility - certainly not worth relying on. The way the data has been collated means it is not like for like. So yes, I think your experiments would be thrown off, but general trends maybe still be identifiable.