Language ability in academic versus real life settings can differ greatly. That's why the CEFR is the perfect tool to measure what students really 'can do'.
As both a teacher and test developer I often have the feeling that the seemingly related fields of teaching and testing are in reality worlds apart. Being active in both fields, I try to mediate, to put it in CEFR terms. On one hand, as a test developer I want to measure skills and collect empirical data on my subjects. On the other hand, I am a compassionate teacher (I think) and I really do want all my students to do well.
Sean McDonald of telc Language Tests explains why not all tests are created equal.