Luxembourg has a small population, about half a million, i.e. this 9/100k means there were only 45 murders. It also has unusual demographics. Unsurprising given the small number, the murder rate appears to fluctuate considerably. Two years after the study (i.e. 2004) the murder rate was 0.4/100k.

Curiously, the figure in the report is quite different from the UNODC rate for 2002 (1.4/100k or 9).

I've just skimmed the report, but from the snippets quoted in the Columbia Public Policy Examiner it seems like a very superficial study, at least where the UK is concerned. For instance, the rise in violet crime in the UK after the handgun ban in the wake of the Dunblane shootings was likely due to other factors: the real question is whether it would have been higher, lower or the same if the ban had not been enacted.

I suspect gun ownership in the UK was relatively low even pre-Dunblane, with registered firearms most dense in rural areas, which tend to have lower violent crime rates in general. So the correlation highlighted by the Examiner is highly unlikely to be causal.