A priori it is difficult to know if a 'blue/purple filter' will or won't affect the detector efficiency. You will have to run some tests yourself to determine exactly what frequencies of EM radiation get absorbed by the filter, and the efficiency of that filter.
If you *really* need to get rid of the blue/purple frequencies then you could probably use an filter which absorbs whichever frequencies you like. If it absorbs UV radiation then your signal-to-noise will go down but then you just count for longer to get better stats.
In any eventuality though if you use a filter you *must* know the attenuation coefficient as a function of frequency over the whole active frequency range. You'll need that to normalise the data.
Alternatively you could just find a material that only absorbs in the UV band. I would think there are some semiconductors that can do that. Aluminium nitride (AlN) I think absorbs preferentially in the UV band because of its 'electronic band structure', but you can check that with the published literature.
To those who understand I extend my hand; To the doubtful I demand to take me as I am.