Actually, this is only true if the CCD has 100% efficiencies -- and that doesn't happen. For instance, typical lab PMT can have their efficiency reduced by several factors. The graph here indicates a maximum of around 30%, although I would swear I have seen manufacturers claiming numbers in the 40s and 50s.

In fact, I doubt a "perfect observation device" for optical phenomena can exist. I'd be interest if anybody had an example of one. Perhaps a Boise-Einstein condensate at low T could detect photons with 1:1 correspondence if the photon decoupled the condensate, but that's total speculation.

Update: I should also point out the difficulties with mono-chromatic filters. These are theoretically possible, but to get a filter down to a one quantum bandwidth... is technically extremely challenging. The detector is the bigger problem for sure, though.