Here's something even more simple. See the attached.
Rlimit is increased to 1 Ohm. You'll take just a small efficiency hit because of this, as with a 250mA load, 62.5mW will be dissipated in Rlimit, while 2,937.5mW will be dissipated in the load. That's still better than 97.8% efficient.
R1 supplies current to pull the gate of M1 high, tending to turn the MOSFET on.
As current begins to flow through the MOSFET, it also flows through Rlimit; as the current increases past 0.63v, Q1's base starts conducting current, which causes the collector to sink current, pulling the gate of M1 back down, which reduces the current flow through the MOSFET, hence through Rlimit.
When C1 becomes fully charged, the voltage across Rlimit drops enough to turn Q1 completely off, and the gate charges to the battery voltage less the drop across Rlimit.
If you have a question, please start a thread/topic. I do not provide gratis assistance via PM nor E-mail, as that would violate the intent of this Board, which is sharing knowledge ... and deprives you of other knowledgeable input.