lsqcurvefit returns solution approximatly the same as initial point

I used lsqcurvefit to estimate the parameters of a complicated model including ODEs and PDE. In order to this aim, the output of the model in the form of time evolution of a variable is compared with the corresponding experimental data. The number of parameters to be estimated are 6. In each trying I used lsqcurvefit, the returned solution is approximately the same as the initial point provided. Also the residuals do not decrease significantly and remain constant while the step size reduces quickly. for example, the residuals for a typical running with 18 iterations remain around 427.76 (which is very large) throughout the iterations but the step size reduces from 10 to 5.72571e-007 and then the optimization terminates. So it appears that the lsqcurvefit does not use a wider trust region and limits only around the initial point. I have tried the different initial points and large values for "DiffMinChange" option field, but none of them made difference. What should I do for this problem? I need that the residuals get at least two order of magnitude smaller. I really appreciate for your helps. Regards