Parallel par stall and low utilization

% Picked 7 starts from the 30 we genrated earlier
Rstarts = [R1; R3; R5; R10; R12; R21; R22];
pathdot = simulation_path;
A=1;
% Lets make pertubations +-5 

cSPSA=3;
a = 0.01;
% set size is decreasing but not the pertubation
cn = cSPSA;
path1 = simulation_path;

seeds = [11934,41900,61321,99410,87252,90467,90188,43116,87667,77541];
% seeds = seeds(6);
maxRuns = 30;
nStart = 7;
Release_final = zeros(length(seeds),nStart,length(Rstarts(1,:)));
PA = zeros(length(seeds),nStart,maxRuns);
for nruns = 4:5
    alpha = 0.602;
    gamma = -0.25;
    % intialize variables
    %In each of the runs assign R to one of the 7 starts 
    R = repmat(Rstarts(nruns,:),length(seeds),1);
    % the orginal SA loop which uses R as the starting point
    deltaJ_plus = zeros(length(seeds), length(R));
    deltaJ_minus = zeros(length(seeds), length(R)*2);
    for n= 1:maxRuns
        an =a/(n)^0.7;
        condition = 30;
        parfor q=1:60
            if q <= condition
                ei = zeros(1, length(R));
                ei(q)=1;
                path2 = strcat('_',num2str(q));
                pathdot = strcat(path1,path2);
                delta_plus = zeros(length(seeds),1);
                for i=1:length(seeds)
                    Rplus = R+(cn'*ei);
                    delta_plus(i) = fullsimcall_GA(Rplus(i,:),pathdot,seeds(i));
                end
                deltaJ_plus(:,q) = delta_plus;
            else
                ei = zeros(1, length(R));
                ei(q-30)=1;
                path2 = strcat('_',num2str(q));
                pathdot = strcat(path1,path2);
                delta_minus = zeros(length(seeds),1);
                for i= 1:length(seeds)
                    Rminus = R-(cn'*ei);
                    delta_minus(i) = fullsimcall_GA(Rminus(i,:),pathdot,seeds(i));
                end
                deltaJ_minus(:,q) = delta_minus;
            end
        end

        % cd back into matlab path (where the fullsimcall.m is located)
        % The problem seems to be that after the parfor is executred we
        % get struck in the simautlion folder and the next function call
        % doesnt get exceuted
        cd(matlab_path);

        deltaJ = ((1/2*cn)').*(deltaJ_plus - deltaJ_minus(:,31:end));
        Rnplus1 = R - an*deltaJ;
        Rnplus1(Rnplus1<=0)=0.05;
        parfor i = 1:length(seeds)
            path2 = strcat('_',num2str(i));
            pathdot = strcat(path1,path2);
            Jnplus1 = fullsimcall_GA(Rnplus1(i,:),pathdot,seeds(i));
            PA(i,nruns,n) = Jnplus1;
        end
        R = Rnplus1;
    end
    Release_final(:,nruns,:) = R;
    cd(matlab_path);
end

I am trying to run this code on matlab parfor and the machine has 72 core. I ran the same code with 100 % utilization before but now the code only achieves 30% utilization and takes 3 times as long as it used to before. The fullsimcall_GA function calls take the input vector and a path and write the contents of the input in a text file in the path specified. Then an .exe file (c++) code is called which uses the input and returns output. Can someone help me out with this?

Thanks,

Karthick