Weird interaction in data structure code for randomized Binary Search Tree

Question

in my code I have created a TreeNode class where each node has a int N field which is the number of subtrees, this is the code for the deletion of a TreeNode in a randomized BTS tree

void remove(int id) 
   {
        if(isID(id)) {   
            removeR(head,id);
        }else 
        {
            System.out.println("There's no warehouse with that ID!");
        }
   }


   private TreeNode removeR(TreeNode h,int id) 
   {
    if (h==null) return null;
    int TreeNodeID=h.getId();
    if(id<TreeNodeID)
        removeR(h.l,id);
    if(id>TreeNodeID)
        removeR(h.r,id);
    if(id==TreeNodeID)
        h=joinLR(h.l,h.r);
    return h;
   }


   private TreeNode joinLR(TreeNode a,TreeNode b) 
   {

    int NA=a.N+b.N;

    if(a==null) 
        return b;
    if(b==null) 
        return a;

    if(Math.random()*NA<1.0*a.N) 
    {
        a.r=joinLR(a.r,b);
        return a;
    } else 
    {
        b.l=joinLR(a,b.l); 
        return b;
    }
   }

If I insert for example a TreeNode with an ID of 1 and select to remove this Node, the code will do : remove(id)->removeR(head,id)->h=joinLR(h.l,h.r), my problem is that in the joinLR function this piece of code

int NA=a.N+b.N;

will make the next two ifs dead code

if(a==null) 
        return b;
    if(b==null) 
        return a;

I did a lot of examples , if I use a or b the code will be unusable , the only solution is to write it after the two ifs , but then the code doesn't work properly , does anyone has any idea?