When using the TreeBagger function in MATLAB to determine variable importance, what is the proper interpretation of OOBPermute​dPredictor​DeltaError > 1?

I have a dataset of 5,000 observations and 10 variables. Each of the observations belongs to 1 of 40 people. When I try to use the TreeBagger function, the OOBPermutedPredictorDeltaError for every variable is greater than 1, which seems incorrect according to my understanding, since this would mean that permuting each variable changes the error rate by more than 100%.

Does this mean that my data is not good or that using TreeBagger won't work? Is there anything that I can change? If all I want is the importance of the variables relative to each other, can I just divide them all by the maximum values to get relative importance? The ultimate goal is to get relative weights of each variable to determine the degree of similarity between pairs of people.

Here is the exact command I am using:

RF_ensemble = TreeBagger(ntrees, zs, labels, 'Method', 'classification', 'OOBPredictorImportance', 'on', 'PredictorSelection', 'curvature');

zs = zscore(myData) labels = 10000x1 vector of person each observation belongs to

OOBE is consistently ~13.5%