how can I make my optimization stable?

Sometimes output pattersearch gives me inverted values that I expected. How can I set it? here are: first my @my_function and second where it is the estimation function.

function f = AV_function_eps(t, rr, lambda)
    %Two different AV node paths (two different refractory periods) with
    %vaiable refractory period
    %Fixed amplitude (suprathreshold) of atrial impulses, no phase 4
    %depolarization 

      tau1 = t(1);
      tau2 = t(2);
      tauemax1 = t(3);
      tauemax2 = t(4);
      alpha = t(5);

      tau1_max=tauemax1+tau1;
      tau2_max=tauemax2+tau2;

      %rr>tau1_max
      p1_2=lambda*exp(-lambda*(tau1_max-tau1)/2).*exp(-lambda.*(rr-tau1_max));
      %rr>tau2_max
      p2_2=lambda*exp(-lambda*(tau2_max-tau2)/2).*exp(-lambda.*(rr-tau2_max));
      %tau1<rr<tau1_max
      p1_1=lambda*(rr-tau1)./(tau1_max-tau1).*exp(-lambda.*(rr-tau1).^2./(2*(tau1_max-tau1)));
      %tau2<rr<tau2_max
      p2_1=lambda*(rr-tau2)./(tau2_max-tau2).*exp(-lambda.*(rr-tau2).^2./(2*(tau2_max-tau2)));
      %rr<tau1
      p1=zeros(size(rr));
      %rr<tau2
      p2=zeros(size(rr));

      x=find(rr>tau1);
      p1(x)=p1_1(x);
      x2=find(rr>tau1_max);
      p1(x2)=p1_2(x2); 

      x=find(rr>tau2);
      p2(x)=p2_1(x);
      x2=find(rr>tau2_max);
      p2(x2)=p2_2(x2);

      % alpha = gamma+.1;
      p_rr=alpha*p1+(1-alpha)*p2;
      f=-sum(log(p_rr+eps));

      function [a, fval, exitflag, OUTPUT] = Estimator(rr, lambda)

Estimation of parameters teta=[t1min t2min tp1 tp2] by jointly maximizing the log-likehood function with respect to teta Algorithm used: PATTERNSEARCH (it is more stable)

        t = [ rand(1) rand(1) rand(1) rand(1) rand(1)]; %Initial Guess
        options = psoptimset('CompleteSearch', 'on', 'MeshRotate', 'off');

        Lb = zeros(1,5);
        Ub = ones(1,5);

        JPF = @(t) AV_function_eps(t, rr, lambda);
        [a, fval, exitflag, OUTPUT] = patternsearch(JPF, t, [], [], [], [], Lb, Ub, options);     %a = Optimum parameters