Spreading one matrix elements to another with weighted random numbers MATLAB

So I was trying to spread one matrix elements, which were generated with poissrnd, to another with using some bigger (wider?) probability function (for example 100 different possibilities with different weights) to plot both of them and see if the fluctuations after spread went down. After seeing it doesn't work right (fluctuations got bigger) I tried to identify what I did wrong on a really simple example. After testing it for a really long time I still can't understand what's wrong. The example goes like this:

1. I generate vector with poissrnd and vector for spreading (filled with zeros at the start)


2. Each element from the poiss vector tells me how many numbers (0.1 of the element value) to generate from the possible options which are: [1,2,3] with corresponding weights [0.2,0.5,0.2]


3. I spread what I got to my another vector on 3 elements: the corresponding (k-th one), one bofore the corresponding one and one after the corresponding one (so for example if k=3, the elements should be spread like this: most should go into 3rd element of another vector, and rest should go to 2nd and 1st element)


4. Plot both 0.1*poiss vector and vector after spreading to compare if fluctuations went down

The way I generate weighted numbers is from this thread: Weighted random numbers in MATLAB

and this is the code I'm using:

    clear all
    clc

    eta=0.1;
    N=200;
    fot=10000000;
    ix=linspace(-100,100,N);
    mn =poissrnd(fot/N, 1, N);
    dataw=zeros(1,N);
    a=1:3;
    w=[.25,.5,.25];

    for k=1:N

       [~,R] = histc(rand(1,eta*mn(1,k)),cumsum([0;w(:)./sum(w)]));
       R = a(R);
       przydz=histc(R,a);
       if (k>1) && (k<N)
               dataw(1,k)=dataw(1,k)+przydz(1,2);
               dataw(1,k-1)=dataw(1,k-1)+przydz(1,1);
               dataw(1,k+1)=dataw(1,k+1)+przydz(1,3);
       elseif k==1
               dataw(1,k)=dataw(1,k)+przydz(1,2);
               dataw(1,N)=dataw(1,N)+przydz(1,1);
               dataw(1,k+1)=dataw(1,k+1)+przydz(1,3);
       else
               dataw(1,k)=dataw(1,k)+przydz(1,2);
               dataw(1,k-1)=dataw(1,k-1)+przydz(1,1);
               dataw(1,1)=dataw(1,1)+przydz(1,3);

       end


    end


    plot(ix,eta*mn,'g',ix,dataw,'r')

The fluctuations are still bigger, and I can't identify what's wrong... Is the method for generating weighted numbers wrong in this situation? Cause it doesn't seem so. The way I'm accumulating data from the first vector seems fine too. Is there another way I could do it (so I could then optimize it for using 'bigger' probability functions)?

Sorry for my terrible English.