Because in Linux they show all "threads" as "processes" (unlike UNIX which shows only processes and requires other tools to look at threads). So what you're actually seeing is multiple "threads". This has to do with the way the Linux community chose to deal with the NPTL (Native Posix Thread Library).
If you had to use the -m it means you, like me, have a server that has an older procps package that didn't do this by default. It doesn't really change anything except the way it displays.
If they are threads they are using the same memory map (/proc/<pid>/maps). When I looked at this in the past I saw it mainly for a couple of Java things. By looking at the process name I was able to identify all the pids (using the -m) then find that there were real "processes" with different maps. All the other "processes" had the same maps as one of the above.
The main reason this becomes important is the way it is viewed would make you think the process is using more CPU and Memory than it really is if you summed up all the "processes". By doing the above exercise I was able to confirm that the memory maps were the same so it was only uses the total for one rather than the total for all.
