The properties of Boolean methods of structural analysis are used to analyze the intern structure of linear or non linear models. Here they are studied on the particular example of qualitative methods of input-output analysis. First, it is shown that these methods generate informational problems like biases when working in money terms instead of percentages, losses of information, increasing of computation time, and so on. Second, considering three ways to do structural analysis, analysis from the inverse matrix, from the direct matrix and from layers (intermediate flow matrices), these methods induce topological problems; the adjacency of the adjacency cannot be defined from the inverse matrix; the binary relation of influence may be non transitive from the direct matrix, with poorer results than with quantitative methods; for layers - based methods, the information carried out by layers is trivial.