When a visitor arrives to the site via a link which contains @piwik_campaign@ (or whatever is set as campaign name), the visit count for the related campaign increases, but the visit count for the site the visitor came from does not. This is true for both referer lists in general and goal details by referer. The lack of real referer data makes it impossible (at least without querying the database manually) to compare the effectiveness of a banner for a certain campaign on different sites.
Campaigns were designed as is, ie. the visitor comes from either a campaign or a website.
It is a bit difficult to count 2 referers for a same visitor as it wasn't designed this way, but this is a good suggestion.
Alternatively, to test the same banner effectiveness on different websites, you could use the 'keyword' for a campaign, the campaign being 'banner A' and the kwd could be the website name that brought the visit from this banner A. This would require that you tag your banner links with the proper campaign name/keyword parameter, as briefly explained in #604