doi .filipos.

don't you think that attitudinals like {.ei} can refer to the speaker only.
In relative clauses it's not possible to use {.ei} refering to anyone described.

What is your opinion about just {bilga} or at least careful usage {of .eida'oi}

Attitudinals express true emotions which you can never be sure about other people.

doi gleki

I think irrealis attitudinals (~propositional indicators) in subordinate clauses should be considered part of them. For one thing, giving short scope to attitudinals in general means they providing equal expressive power to main bridi and abstraction bridi. More importantly, for the particular case of irrealis, I see no general useful alternative interpretation.

So,
{ko'a ca .ei cliva} "ko'a should leave now"
{lo du'u ko'a ca .ei cliva} "The proposition <koha should leave now>"
{lo se du'u ko'a ca .ei cliva} "A sentence expressing that koha should leave now"

You could use {bilga} in an alternative translation. {da'oi} does not help if you think that {.ei} has long scope.

Irrealis attitudinals are a fundamental part of what is being stated, not a side emotional expression like {.ui}.