I'd guess that the transparency of the water is the cause of your low output signal. As to the drift, it might be that the current flowing through your receiver with no interrupting water is causing the LED to increase conduction and so drifting its output.
Have you tried reducing the 100k resistor down to say 10k.
You might get better results by modulating the transmitter LED signal to a pulsed, rather than fixed DC.
Alternative detection could be obtained by letting the drops fall onto a pair of contacts. The droplet will cause slight conduction before it drops off the probe tips and this conduction should be easily detectable.