The purpose of this note is threefold. 1) It draws the readers' attention to some papers addressing initial state learning not cited in the aforementioned paper. 2) It points out that the class of control systems addressed in the aforementioned paper is restricted to that of systems with state-space dimension equal to output space dimension, but that this hypothesis can be removed. 3) It analyzes the point of convergence of the algorithm proposed in the aforementioned paper when applied to systems with internal dynamics.