The Least-Squares Progressive-Iterative Approximation (LSPIA) method offers a robust and efficient approach for data fitting. Non-Uniform Rational B-splines (NURBS) are typically used as approximation functions due to their powerful shape representation. However, the traditional LSPIA application in NURBS comes with the restriction that only control points are adjusted to approximate the given data points, with weights and knots remaining fixed. To enhance fitting precision and overcome this constraint, we present Full-LSPIA, an innovative LSPIA method that jointly optimizes weights and knots alongside control points adjustments for superior NURBS curves and surfaces creation. We achieve this by constructing an objective function that incorporates control points, weights, and knots as variables, and solving the resultant optimization problem. Additionally, to construct lightweight geometry, we present a knot removal strategy named Decremental Full-LSPIA based on Full-LSPIA. This strategy adaptively removes the redundant knots within a specified threshold and determines optimal knot locations. The proposed approaches maximize the strengths of LSPIA. Compared to the classical LSPIA method, Full-LSPIA offers greater fitting quality for NURBS curves and surfaces while maintaining the same number of control points. Moreover, Decremental Full-LSPIA yields fitting results with fewer knots while maintaining the same error tolerance.