Statement A is correct. Although screening ignores all information in alphas except the rankings, it often mitigates transaction costs through the judicious choice of the buy, sell, and hold lists (e.g., buffering rules to prevent high turnover). Statement B is incorrect because stratification controls risk by matching the benchmark's risk dimensions (matching the weights of categories to the benchmark), not by overweighting low-risk categories. Statement C is incorrect because it is the stratification technique that requires dimensions to distinctly and exclusively partition the stocks, whereas the linear programming approach does not have this strict partitioning requirement. Statement D is incorrect because quadratic programming requires many more inputs (a full covariance matrix), which maximizes (rather than minimizes) the impact of noise and the potential for poor calibration.