Dear Mamado,
I came accross ur case by accident and I would like to help, hope I'll be able to.
Let us agree first that solutions for those problems arise from suggesting theories of origin of the problem, just as ur team suggested preliminary checking the curent transformers so, what I'm saying that I'm presinting a theory that might be 100% out of realism.
Well, to limit the problem, i will say that that the same problem would occured if u were using 50N or 51N ( same protection without the directional portion) and that's because ur problem is in the accuracy of measurment of the current regardless of the Voltage part.
My theory is that the amount of current that you are reading is three times the victorial sum of all the ( third and multiple of third) harmonics. that's because the rule states that:
The amount of the residual current ( I1 + I2 + I3 ) in the abcense of an earth fault = 3 x victorial sum of all the ( third and multiple of third) harmonics.
Clear rules of this type of protection states that u have to make ur protection insensitive to those types of currents.
Now that you have a vision for your root cause, u may think of work arrounds as to have a new tripping pattern taking into consideration the third and x third harmonic contents or filtering the sensed current from those components.
hope that helps and sorry for the language as I'm suffering from a very bad arabic typing.