Harmonic Closeness for disconnected Graphs

[tigger]

Hello everybody,

I wanted to measure the closeness but realized that my network is disconnected so that I have more than one component and the closeness centrality is not really "true". I was really pleased when I realized that there is another measure called harmonic closeness that deals with the same issue for disconnected networks (in detail: https://infoscience.epfl.ch/record/2005 ... ASNA09.pdf?) and gephi can measure it. But my results can´t be right. I tried measuring it with some easy networks designed by my own and I have completely different results.
It seems that gephi is normalizing the harmonic closeness centrality with the number N of nodes of the (disconnected) component. But officially it should be normalized with the number of the nodes of the whole network.
Is it a known problem?