sqp-minimization problem

[Urs Hackstein]

Dear list,

I encountered the following strange behaviour of a sqp-minimization
problem. Let a function minimize11version3h defined as follows:

function mini=minimize11version3h(b)

funp11wrap = @(x)funp11version3(x,[real(b),imag(b)])
y0  = 5*ones(9,1);
lbound = [0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,5]';
ubound = [20,20,20,20,20,20,20,20,200]';
h=@(x)(1.85*x(1:4,1)-x(5:8,1));
mini=sqp(y0,funp11wrap,[],h,lbound,ubound,500)

function erg=funp11version3(x,p)
erg=(p(1)-real(dominant11(roots(polyversion3(x(1),x(2),x(3),x(4),x(5),x(6),x(7),x(8),x(9)))))).^2+(p(2)-imag(dominant11(roots(polyversion3(x(1),x(2),x(3),x(4),x(5),x(6),x(7),x(8),x(9)))))).^2
endfunction

Here polyversion3 creates a polynomial in the variables x(1),...,x(9)
and dominant11 selects the root with the greatest real part.
Now minimize11version3h(-3.1129e+10+1.2570e+11i) gives erg=...*10^21.

However, if I use the following function called
minimize11version3(-3.1129e+10+1.2570e+11i), I receive erg=1.7156e+11
and mini=(20,20,20,1.4311,3.5427,4.4572,20,2.1468.197.3116)

function mini=minimize11version3(b)

funp11wrap = @(x)funp11version3(x,[real(b),imag(b)])
y0  = 5*ones(9,1);
lbound = [0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,5]';
ubound = [20,20,20,20,20,20,20,20,200]';
mini=sqp (y0,funp11wrap,[],[],lbound,ubound,500)

But this solution achieves obviously the condition h\geq 0; thus it is
also a better solution to the minimization problem
minimize11version3h(-3.1129e+10+1.2570e+11i) than the solution found
by minimize11version3h.

Could someone explain these results of the sqp-minimization problems to me?
Thanks a lot in advance!

Best regards,

Urs Hackstein