This paper deals with the optimization of a set of transmitted sequences and their associated mismatched filters, for the coherent MIMO radar. A minimization problem may be considered , with an objective function that measures for instance the correlations within the set. In the literature, it is usually solved alternatively, meaning that the optimization is performed on the sequences while the mismatched filters are set, and vice-versa. In this paper, an iterative method that proceeds jointly is introduced. Sequences are derived with a gradient descent, and mismatched filters are chosen in such a way that they are optimal in the ISL (Integrated Sidelobe Level) sense. Simulations show interesting results, as a joint optimization performs better than a separate one, even with relatively short sequences.