In some of the data file from NCBI SRA, the fastq formats are being messed up by including spaces in the seqname fields. If this happens, then the part after the space plus the next line up to the max length is taken to be the quality scores and fastq2bfq will not complain unless one of the qualities is '@'. In that case, it says that there is an "Inconsistant sequence name" and continues parsing the fastq file incorrectly.
To verify this, create a fastq file with spaces in the seqname, run fastq2bfq and then bfq2fastq. There will not be any complaints from MAQ but the bfq file is totaly messed up.
I think (but have not tested and recompiled MAQ myself) that the problem is that while reading the second seqname (at line 86 of seq.c) after coming across a space, it fails to go to the next '\n' like it does on line 69 of the same file. If there is another check to make sure that it is at an '\n', this should at least allow MAQ to create a proper bfq file.