Hi @styfen.schaer,
Thank you so much for your quick reply!
I have produced a reduced version of the code that reproduces the error I described above.
I zipped it and I upload it through this reply.
Inv_TC1_1_for_dbg.zip (9.7 KB)
You will find the reduced example in the jupyter notebook “Inv_TC1_1_for_dbg.ipynb”
My PCE models are replaced by models which return (1,1,…,1), to (5,5,…,5), the same size as the original models output vectors of my problem and original data are replaced by these constant vectors.
Furthermore, in the meantime I have thought of a possible answer to my problem and tried transposing the MOMap array in order to get a structure that looks like :
[ [1, 1, …, 1],
[1, 2, …, 66] ]
Rather than : [[1,1], [1,2], …, [1,66]]
This was inspired by this post (Bayesian inference with multiple forward models, each one with different vector-valued outputs) which is about UQLab (Matlab) but the MOMap seems to be a transposed version of mine (which I defined based on UQpyLab manual : " Nested list of Nout,g lists of length 2 and type integers").
This is implemented in the jupyter Notebook “Inv_TC1_1_for_dbg_transpose.ipynb”.
With this version, I don’t get the error “MOMap of data group 1 is not consistent with supplied data” anymore, the analysis is running. Could there be an error in UQ[py]Lab manual ?
Though, with this version I get another error after intermediate computations : “Error: Index in position 2 exceeds array bounds. Index must not exceed 1.”
This last error might be due to inconsistancies in the dummy functions and data that I have quiclky defined for debugging purpose.
I try running this solution with my original problem and will come back to you if it works.
Thanks for having a look into my problem ! 
Best,
Guillaume Gru