TreeBagger removes elements from the dataset during training process

I am using the TreeBagger for classificate my data. I would like to extract some stats about it, like confusion matrix, but I can't due to a mismatch between the training sets.

%That's how I set the TreeBagger: 
b = TreeBagger(50, X, y, 'method', 'classification', 'NVarToSample', 10, 'oobpred', 'on',   'MinLeaf', 10);

%output:
b = 

   Ensemble with 50 bagged decision trees:
               Training X:            [1801x32]
               Training Y:             [1801x1]
                   Method:       classification
                    Nvars:                   32
             NVarToSample:                   10
                  MinLeaf:                   10
                    FBoot:                    1
    SampleWithReplacement:                    1
     ComputeOOBPrediction:                    1
         ComputeOOBVarImp:                    0
                Proximity:                   []
                    Prune:                    0
              MergeLeaves:                    0
                 TreeArgs:
               ClassNames:            '-1'             '1'

%Note: y was originally composed by 1834 elements

[y1, scores] = oobPredict(b);
[conf, order] = confusionmat(y, y1)

%ERROR: y1 is just 1801 elements long!

Why does it happens?