Ok, I also posted this on the Google Analytics support forum. You'll find a lot of screenshots there about my problem.
The bottomline is that the original homepage in a A/B test receives far more 'new visits' then the variant. This seems to be the main cause for the orignal homepage to be the winner in the experiment.
Has anyone seen this before? What could be the reason for the original page to receive more new visits? And shouldn't homepage A/B tests results not always be filtered by visits that only saw either the homepage or the variant (so filter on only 'new visits' that occured during the A/B test)?
Any input is much appreciated!