Early stages modelling of processes involves issues of classification of variables into inputs, outputs and internal variables, referred to as Model Orientation Problem (MOP) which may be addressed on state space implicit, or matrix pencil descriptions. Defining orientation is equivalent to producing state space models of the regular or singular type. Studying the conditions, under which such models may be derived, as well as the structural properties of the resulting oriented problems, is one of the issues considered here. Oriented models of the S(A,B,C,D) type have transfer functions which may not be well behaving as far as properties of input, output regularity and nondegeneracy. Defining subsystems by reduction of the number of inputs, outputs such that the resulting transfer function is well behaved as far as preservation of degeneracy, is a problem considered in the context of systems having physical input, output variables, which have to be preserved.