AnalogKid’s point about the recharge path is the first thing I’d address. Adding another capacitor stage is possible in principle, but it needs a drive waveform that can charge and reset the flying capacitor on successive cycles. The PNP’s strong pull-up alone doesn’t provide that once the LED is moved downstream.
A pull-down resistor could provide a return path, but its value would affect recharge time, battery consumption and potentially the oscillator’s operation. Simply adding two diodes and another capacitor doesn’t resolve those issues.
If you want to explore a capacitor-only solution, I’d separate the voltage multiplier from the flash timing: use an oscillator and output driver capable of sourcing and sinking current at the battery voltage, charge a reservoir capacitor, then switch a current-limited LED pulse from it. That lets the pump replenish the stored energy between flashes. An ideal doubler from 1.5 V only gives 3 V before losses, so there may be very little headroom for a blue or white LED at the desired current.
If two cells are acceptable, Talking Electronics already includes a
“3v WHITE LED FLASHER” in its
transistor circuits collection. That would be a useful capacitor-boost circuit to study before designing another stage.
For comparison, the
Engineering Mindset Joule-thief project illustrates the inductive approach mentioned earlier. It powers a white LED from a low-voltage cell, but would need additional control to produce distinct, slow flashes.
Does the design have to stay with one 1.5 V cell, and roughly how long should each flash last? Those constraints would help determine whether another capacitor stage is practical.