We do have to recognize the limitations of the sample: looking at the June 2011, American Rifleman, the armed citizen, as an example, most are clippings from newspapers or television reports. The NRA does ask for first hand accounts, but I suspect NRA asks documentation.
Due to bias in the news about guns, positive gun stories are exceptional. Of self defense incidents recounted to me, half were not reported to the police and none generated news coverage.
That does not mean that this sample is not useful, but we must keep in mind that it is primarily a sample of successful self defense incidents that were deemed newsworthy.
You won't be able to project absolute numbers about self defense with a gun, but you do get some good stats on the nature of civilian self defense from the sample as long as you recognize it is mostly a sample of successful self defense incidents deemed newsworthy by newspapers or television.
It is kinda like using the FBI UCR table on justifiable homicides: killing of a felon in the line of duty by a police officer, and killing of a felon during commission of felony by a citizen (from police report, not reflecting eventual adjudication by coroner, medical examiner, district attorney, grand jury, trial judge, trial jury or appellate court). You can get a good idea of methods or weapons commonly used, but not the absolute number of events per year.