In the Generalized Assignment Problem, It requires to minimize the cost of assignment of ‘n’ tasks to a subset of ‘m’ agents. Each task is to be assigned to one agent and each agent is limited only a maximum amount of resources available to him. A Lexi Search algorithm using Pattern Recognition was developed and presented already by the authors, along with computational experience. In the present paper, fresh computational experience is reported along with the refinements made in the algortihm.