Although I cannot send DDLs they allowed me to run and post the result for the "missing stats" script proposed by Abdelhak and I attach it along with the plan when I used "parallel" hints. I shall also try to decipher it.
Concerning the temporary tables solution, it may seem strange but the decision not to use it is not mine, so I didn't focus on it. All these suggestions are very interesting and helpful and they you a lot for your effort. Also Jeff said that parallel run is meaningless for a query returning a few rows, here it returns only 35 rows. Finally the idea for
query 1
union
query 2
union all
query 3
union
query 4
fits in our case, I mean that it is true that the second union can be replaced by union all, but again it had no effect on execution duration.
Thank you all a lot for your valuable help.