Thank you for your report! I recently faced the same problem and concluded that you were on the right track. Since the FASTQ standard seems to state that comments can appear after the '+', these comments (which are common in NCBI's data) need to be removed from the buffer.

The while loop on line 86 will read up to the first whitespace. What is left needs to be taken and a line similar to what is on line 69 should do the trick.

The result is that the quality scores can become "out of sync" and if one of the qualities is not '@', then the problem will go unnoticed by the user.

I am uploading a patch now that hopefully solves this problem -- it does basically what you suggested. Thank you!