Arnoldi method to find eigenvalues

I'm trying to implement the Arnoldi method with the inverse power method to find eigenvalues of large pencil matrix.

I have followed the practical implementation in Saad's book and I started with a small matrix to check if the code work well. As I understand H is a square matrix and has size of the number of the iterations but the resulted H is of size 3x2 and V is 4x3. I'm not sure if this is correct and I do'nt know how I can find the eigenvalues of H and the corresponding eigenvectors. Here is my attempt, and I really appreciate any help..

A=[1 0 0 0;0 17 0 0;0 0 -10 0 ;0 0 0 5]
n=length(A)
B=eye(n)
a=0  %interval [a,b] 
b=10
m=2;  %iterations
x=ones(n,1); %constant vector
shift=0.5;
V = zeros(n,m);  %orthonormal basis of Krylov space
V(:,1) = x/norm(x);
H = zeros(n,m);  %upper Hessenberg matrix
C=A-shift*B; 
[L, U] = lu(C, 'vector');
for j=1:m     
    w=U\(L\(B*V(:,j)));
    for i=1:j    %Gram-schmidt 
        H(i,j) = V(:,i)'*w;
        w = w - V(:,i)*H(i,j);
    end
        H(j+1,j) = norm(w,2);
       V(:,j+1) = w/H(j+1,j); 
end