Hi barthess,
thanks for this precious feedback. I am working on a restyle of the current implementations. We are far from the definitive version and the idea is to improve driver until we have something suitable for most common uses.
Note that bias and sensitivity are two abstract concept related to a linear correction according to formula
Sens * Data + Bias = Cooked
Whereas Data is a 1xn vector, Sens is a nxn matrix and Bias is a 1xn vector. Since cross sensitivity values are almost 0 in our model we are considering Sensitivity like a nx1 vector.
By the way:
The second point... My bad! I have to improve the whole error management. Thanks for notice me.