What is the added value of hessian pattern if the hessian is user supplied

Hello,

I am trying to understand how adding hessian pattern helps in the case where we have user supplied hessian functions. Isn't it the case that hessian function imposes some sparsity itself? Same question goes for the gradient. I would appreciate it if you can direct me to related literature or help me understand if it has a relatively simpler explanation. Thank you!