count performance

Hi the experts !

Actually, i have a bad performance with this counting request :

count(
for $a in input()/suiviPublication
let $b := $a/publication[@archive=“N”]/diffusions/diffusion[@format=“LASER”]
where $a
and ( $b/diffusions/diffusion/@etat=“ERREUR” or $b/diffusions/diffusion/@etat=“ENCOURS” )
return $a
)

For information, i tried the compound index with the 3 attributes
(@archive, @format, @etat) without success.
So i put a standard index on each one of these 3 attributes
. But i also had the same performance.
the disadvantage: i can’t change the schema’s structure, just putting
index.

Can i have any help ? a solution ?

Thanks

Tatoum