Previous research proved that teaching spreadsheeting from a programming perspective is much more effective than the widely accepted tool-centered surface approach methods. Spreadsheeting as an introductory programming approach allows students to build up schemata leading to contextualized, concept-based problem-solving. Furthermore, it provides tools for real-world problem-solving in other disciplines, and supports knowledge-transfer to database management and "serious" programming. The present study provides the details of a nationwide testing of Grades 7-10 students on how they evaluate their spreadsheet knowledge, which classroom activities form their self-assessment values, and the results of three spreadsheet tasks evaluated by the SOLO categories of understanding. The comparison reveals that most students' spreadsheet knowledge is pre-structural. On the other hand, they assess themselves much higher, which is primarily based on the number of activities carried out in classes. Traces of conscious problem-solving and knowledge-transfer within the scope of spreadsheeting are hardly detectable, while knowledge brought from mathematics is recognizable. In general, we found proof that the pieces of knowledge remain unconnected, not allowing students to reach the relational level of understanding and build up long-lasting knowledge.