How can I use lsqnonlin in an optimization procedure where it is possible that the evaluation of the objective function might return NaN?

I have read that lsqnonlin optimization solver can work even if the objective function returns NaN or Inf at some point within the design space. I understood that lsqnonlin would not take this solution into account and continue the optimization procedure with valid points.

However, I am getting the following error:

??? Error using ==> eig Input to EIG must not contain NaN or Inf.

Error in ==> trust at 28 [V,D] = eig(H);

Error in ==> trdog at 110 [st,qpval,po,fcnt,lambda] = trust(rhs,MM,delta);

Error in ==> snls at 365 [sx,snod,qp,posdef,pcgit,Z] = trdog(x,g,A,D,delta,dv,...

Error in ==> lsqncommon at 149 [xC,FVAL,LAMBDA,JACOB,EXITFLAG,OUTPUT,msgData]=...

Error in ==> lsqnonlin at 241 [xCurrent,Resnorm,FVAL,EXITFLAG,OUTPUT,LAMBDA,JACOB] = ...

Error in ==> main at 53 [x,resnorm,residual,exitflag,output,lambda,jacobian] = lsqnonlin(@CostFuncLsqnonlin,x0,lb,ub,options);

I have tried a lot of possible solutions but haven't been able to solve the issue. I would really appreciate some assistance.

Thanks a lot.