Why convictions are few when FIRs are many
The victim is a player too. How long a complainant stays in the case decides who is left to testify, which decides the conviction probability the accused actually faces, which decides whether pleading or waiting is cheaper.
Who is still there when the trial happens
A complainant, and the witnesses who come with them, stays in a case while the discounted value of a conviction (with any compensation) exceeds the cost of attending for the expected wait plus the private compromise on offer. The curve is the share who persist against the wait; the marker is the current scenario.
The funnel: 10,000 FIRs under the current scenario
Where cases die. Convictions are few not because trials acquit but because most cases never reach a judgment: the police close some, the complainant gives up on more, and the accused buys the rest.
The two loops
The first loop runs through the accused: a long wait makes pleas rare, rare pleas keep the trial list full, a full list keeps the wait long. The second runs through the victim and is perverse: a long wait makes complainants quit, quitting relieves the docket, so the system settles at a high wait with the clearance done by attrition rather than judgment. Adding judges alone shortens the wait a little, fewer victims quit, more cases survive to trial, and the wait climbs back. The reform works on both loops at once: a fresh-case list keeps the complainant in, Item A1 gives them a reason to stay, and Items E1–E2 mean that even a complainant who gives up leaves usable evidence behind, so waiting the victim out stops paying.