Identification of the cointegration matrix in the vector error correction model

I am trying to estimate a vecm model on a time series of dimension 3.

This is what I do

load Data_Canada
Y  = Data(:,3:end);
[h,pValue,stat,cValue,mles] = jcitest(Y);

************************
Results Summary (Test 1)

Data: Y
Effective sample size: 40
Model: H1
Lags: 0
Statistic: trace
Significance level: 0.05


r  h  stat      cValue   pValue   eigVal   
----------------------------------------
0  1  37.6886   29.7976  0.0050   0.4101  
1  1  16.5770   15.4948  0.0343   0.2842  
2  0  3.2003    3.8415   0.0737   0.0769 

By looking at the Johansen test I derive that there are 2 co-integration vector. So I proceed to estimate the model using the code:

Mdl    = vecm(3,2,2);
EstMdl = estimate(Mdl,Y);

According to the matlab documentation the co-integration matrix is

>> EstMdl.Cointegration

ans =

    1.2305    1.4689
   -5.5932   -2.2364
    4.1854    0.8046

What I cannot understand is which normalization is used by default. The co-integration matrix is not uniquely identified, so there should be some hidden assumptions that I cannot find in the documentation.