Yes, if you use a silicon solar cell then a 950nm LED would be better than white or blue.
As long as the photons have enough energy to be received by the solar cell, any extra energy in the photons (more blue-ness) is wasted, yet it requires more voltage drop in the LED to create these bluer photons.
So a 950nm LED and a silicon solar cell or photodiode would be a good combination, which you can even buy in a DIP package. Some of them have many photodiodes in series as they are intended for driving the gates of MOSFETs, and can produce > 5 Volts.
This one can provide 7 Volts open-circuit and 12 or 20uA short-circuit:
At the maximum-power point, it will produce slightly less than the short-circuit current, but if you shunt-regulate it to significantly lower voltage e.g. 3.3V or maybe even 5V then it will produce pretty much the short-circuit current.
The efficiency is not good, but it is not as bad as it seems from just looking at the current in and out, since the input voltage is 1.4V so it is stepping up the voltage.
If you don't want to use one of those opto-couplers and you don't want to use many photodiodes in series, and if you get to control the blinking of the LED, then it might make sense to modulate the LED at ~
100kHz, ~ 30% duty cycle, and feed the output from the photodiode straight into a step-up transformer or flyback inductor, and then rectify it. This would avoid any losses in a switching transistor (let alone gate drive generation) in the circuit connected to the photodiode that only provides about 0.6VChris