Sometimes the heuristic to guess the source code line and function name for a loop is wrong, and in that case it's quite hard to work out the truth given the current output. Ideally the csv should contain more information besides the heuristic (e.g. the symbol name) to mitigate this. Also ideally the heuristic is improved.