Update: GDF Writer process
The GDF Writer Box writes data in the file as it arrives. This makes it difficult to fill the header information.
Notably the PHYmin/max and DIGmin/max which are used to scale/offset the data and represent the signal.
It's only possible to determine their value once all the signal has been received, in order to write all the data with the the resulting scale and offset.
At the moment, since the values are encoded on 8 bytes, which does not really require for scaling/offset, the box's PHYmin/max and DIGmin/max have been hardcoded to the largest floating point and integer possible, generating a scale of 1 and offset of 0.
The generated file is not getting read properly in EEGLab, and it seems to be because of the PHYmin/max and DIGmin/max which are not representative of the signal.