Hello Rolando,
normally neither free variables (variables lb==-inf and ub==inf) nor dense columns (columns of the constraint matrix A with "many" nonzero elements) present a problem that you should be worried about too much. We include these two pieces of information in the barrier statistics output only for informational purpose, and to aid debugging unusual performance behaviour of the barrier algorithm.
The background of these two particularities is related to the algorithmic core of the barrier algorithm, and it is rather technical. In brief, the KKT system stemming from primal, dual, and complementarity considerations suggests solving a linear system with the matrix A*A'. Now if A has a dense column, the latter matrix will have a dense submatrix, which makes solving the linear system costly. Free variables, make it difficult to assert positive definiteness of A*A'.
Gurobi has some algorithmic machinery in place to deal automatically with dense columns and free variables. If you observe unsusual performance patterns of the barrier algorithm, the shown statistics on free variables and dense columns are simply gentle reminders that these could play a role.
As said initially: usually you won't have to worry about them!
Robert