Hi Andrei,
I tried the perform the same , got better performance by creating a constant column on both the input of the join nodes .
But when I analysed the explain plan for the same both the the views provided me with the same statistics.
Also when I tried to utilize the view as input to another view over here again the view created by having join with constant column outnumbered the performance of the view with join without the column .
In the explain plan both the joins are shown to be acting as inner join .
So I believe when it would be executed for larger dataset we can hope to find some variation in the performance of the views.
At least in this case we do not have any documentary proof to buttress our case .So would recommend you to go ahead with the one which is providing you better performance .
Do share your results in the forum . Best of luck 
Regards
Rohit