The first 20 bases are continuously mapped to your genome. Suppose your read started at 1002, then, let's see how the read is for this cigar string. Suppose your cigar was 20M 1D 20M 60N 30M 1I 20M 100N 9M => read length = sum of all M/I/S operations. Note that still there might be mismatches. Your cigar of 100M indicates that the read starts at 1002 and is mapped continuously for the next 100 bases. I is insertion into the reference and D is deletion from the reference. The CIGAR string might require some explanation. POS: 1002 - position on the genome where the first base of this read matches (refer to the documentation reg. Type in your flag number and it explains what they mean. To start with therefore, use this link to interpret them.

Probably hard to grasp for biologists at the beginning. FLAG: 163 - read paired, read mapped in proper pair, its mate maps to reverse strand, this read is second in pair. In case of paired end reads, you'll have them twice, assuming both reads of the pair are mapped. Since you also asked for explanation regarding the different columns in your read: 1. However, there might be conflicting options in this regard, I believe. If you're doing a gene expression analysis or anything that deals with count data, then probably not a good idea. I'd suggest to do so, if you're calling SNPs. That being said, it depends on what is your downstream analysis as to whether its essential to remove duplicates or not.

You can remove duplicate reads with picard tools' MarkDuplicates (link from Jorge) as follows: java -jar MarkDuplicates.jar INPUT=my_file_input.bam OUTPUT=my_file_output.bam METRICS_FILE=my_metrics.txt REMOVE_DUPLICATES=TRUE TMP_DIR=my_tmp_dir

Bam file format