Hi Abdallah,
The mode combiner basically strips away the spatial information of a signal and leaves you with a time domain signal. You could then run the signal through a Multimode Generator to attach the spatial information you want (Convert the modes).
The problem you are facing with the eye diagram visualizer is that there are m-ary threshold detectors in these project files, which depending on your defined threshold will assign a m-ary value to each bit. Using this sequence to drive a NRZ Generator results in a “perfect” signal with no noise or variation, this is what accounts for the eye diagram appearing as a single line instead of many. This is also the reason why the BER and Q factor are giving you unrealistic results.
Hope this helps!