For solving constrained multi-objective optimization problems, an evolutionary algorithm using the two-stage non-dominated sorting and the directed mating (TNSDM) has been proposed. To generate offspring, the directed mating utilizes useful infeasible solutions having better objective values than feasible solutions in the population while conventional approaches avoid to use infeasible solutions as parents. Actually, the directed mating significantly contributes to improving the search performance of evolutionary constrained multi-objective optimization. To cross genes (variables) of selected parents, so far commonly-used crossover operators have been combined with the directed mating. To further improve the effectiveness of the directed mating in continuous problems, in this work we propose a method to control the amount of variables inherited from useful infeasible parents by varying the variable exchange probability in the SBX-based variation. Experimental results using two benchmark problems, TNK and mCDTLZ, with 2-4 objectives show that the effectiveness of the directed mating in continuous problems is further improved by increasing variables inherited from useful infeasible parents.