d(y,A) - d(x,A) \geq -d(x,y);
Does not imply
-d(x,y) \leq -(-d(x,A)+d(y,A))
but the transition between the last two lines of your proof relies on it. You just switched the sides and direction of the inequality (which is fine) and multiplied one side by -1 (not good). If we could do...