Your logic is not clear from the example you've posted. I am assuming that you want to send the records which DO NOT have a NULL in column 1 in the output and that you don't want duplicates on col2 in the output.
You can read the file through a single sequential file stage (by adding two File properties). After this you can sort on col2 and then in a downstream transformer you can send the first row with a NOT NULL col1 and then drop the other rows.