It is a warning, not an error. Using too low tolerance for total energy calculations might however give rise to large errors in Dynamical matrix, and your tolerance parameter is indeed pretty large, 10^-3, which is even larger than default value of 10^-4. I would set it to 10^-5. In fact, one should check if Dynamical matrix does not significantly change when going, e.g., from tolerance of 10^-3 to 10^-4. If it does, one may even consider using 10^-5 to see if tolerance 10^-4 gives you converged Dynamical matrix.