Hi, Miriam, interesting question. I think that the major ESL tests (TOEFL, IELTS) are about as good as possible in the real world, but still fall far short. Here are my two main objections: The tests put a lot of emphasis on essay writing, oral presentation (rather than natural conversation) and other skills that are more or less language-independent. If students are good at taking tests, writing essays, and public speaking in their native language, they’ll do well. If on the other they’re bad at that in their native language, they’re at a serious disadvantage on the exam. Secondly, a native speaker masters basic grammar, then intermediate, and finally advanced. Many native speakers would fail to score higher than B2 for this reason — the advanced grammar just isn’t that relevant for most native speakers. A non-native, on the other hand, tends to skip steps, and might end up with a 75% mastery of the grammar at each level, thus also scoring a B2 on the test. The problem is that in real life, this test-taker’s 75% mastery of little-used advanced grammar does not make up for the 25% shortfall at the all-important basic and intermediate levels. Of course, the very fact that these tests are flawed is what makes targeted preparation so important. 

[2/2]

So, I think the Russian exam would give far more false negatives than the French exam (a false negative being someone who clearly has the necessary level, but still failed the exam). On the other hand, the French exam would give you lots of false positives (people who don't have the level, but still passed the exam). In particular, to pass the DALF C1, you need 50% overall, and at least 20% in each of the four subsections. Now, if you're a poor reader, you cannot pass (because three of the four parts entail reading relatively complex texts). If you're a poor speaker (far weaker than C1), however, you absolutely can pass, as long as your other three skills are good enough. For the TRKI-3, you need 66% in each of the five subsections (actually, I've seen one source that claims that 60% is acceptable in one of the five subsections, and you need 66% in the other four; information for the TRKI exams is all over the place...), and so you cannot pass if you have a truly deficient skill. 

What does all that mean? You tell me. I still think the exams are far more accurate than self-evaluation. 

[1/2]

Here's what I think. Language proficiency exams are a somewhat blunt instrument for ranking people in terms of language proficiency. But, blunt as they are, they are still far more accurate than self-evaluation, especially when you need to compare levels of people whom you don't know well (so, you don't know if they love're timid or love to brag or what have you), which is precisely what happens when you need to hire someone for a job or decide whom to admit to a university. 

Do they correspond to the CEFR scale? Well... For one thing, some tests demand a lot more test-specific preparation than others. So, a person who obviously has the necessary level might fail simply because s/he failed to sufficiently practice the very particular tasks that the exam tests. I took (and passed) two C1 level exams in the past nine months: DALF C1 (French) and TRKI-3 (Russian). Let me tell you, the Russian exam is much harder than the French exam, and it requires much more test-specific preparation. If our friend K P presented himself at the TRKI-3 exam after, say, a couple of hours of preparation (to familiarize himself with the format), he'd royally flunk the speaking and writing parts. You really need to know (and practice!!) what the test makers are after. K P's French equivalent would certainly pass the DALF C1 after a couple of hours of preparation (to familiarize himself with the format).