Yes, you're right.
>If so, I need to specify ISTATE in $MCSCF, right?
Yes. I forgot to mention this, sorry.
Also, I forgot to mention that after the state-averaged geometry optimizations with WSTATE(1)=0.1,1 and WSTATE(1)=1,0.1 you should do single-point state averaged calculations with WSTATE(1)=1,1 in the optimized geometries to obtain well-balanced orbitals equally good for both states.
>Why 0.1 is used in both cases?
Well, you may use any small weight of the ground state in the optimization of the excited state, and the same weight of the excited state in the optimization of the ground state.
>For first triplet excited state, do I still set WSTATE(1)=0.1,1?
If you are optimizing the lowest triplet with CISTEP=GUGA and MULT=3 or ALDET with PURES=.t. and MULT=3, it is not necessary, because this triplet state is the lowest one, and it will not mix with the singlet ground state.
However, if MULT=1, CISTEP=ALDET and PURES=.f., you'll have singlet ground state and triplet excited state which are allowed to mix (PURES=.f.). In this case, you should set WSTATE(1)=0.1,1 for triplet.
>For higher excited state such as the second singlet and triplet states, do I set WSTATE=0.1,0.1,1?
If excited states are allowed to mix (say, two excited singlets without symmetry restrictions), you should set small weights to all states below the state of interest. In your example, WSTATE(1)=0.1,0.1,1 is OK.
And don't forget to perform single-point state-averaged calculation in the optimized geometry with all weights of the states of interest set to 1. In different geometries, only energies obtained this way may be compared.