I think dividing the number of citations by the number of publications would be better. Since th journal impact factor has nothing to do with a publication impact!!
This impact factor is not the impact of my article. It is the impact of the journal due to many parameters concerning the journal itself and how the journal manage its business. It has nothing to do with whether or not my article is good. There are many articles in such journals that haven't been sited even once.
Actually I have never seen a "impact per publication" parameter on ReserchGate calculated the way you describe. Do you have this in your profile currently?
There is a "Average Citations per Article" parameter in ThomsonReuters/ISI/WebOfScience that is calculated as you suggest would be more fair.
1. Impact factors were introduced to score the importance of Journals. Not to be applied to calculate the score of their contributors;
2. If you want to calculate the importance of a scientist or researcher, shouldn't you include as an important (most important ) factor of that scale, every book publication with ISBN (that does not weigh in our impact factor calculation, even if it is a bestseller...)???