As part of the BIP66 rules, once 950 of the last 1,000 blocks were version 3 (v3) blocks, all upgraded miners would reject version 2 (v2) blocks.
Early morning UTC on 4 July 2015, the 950/1000 (95%) threshold was reached. Shortly thereafter, a small miner (part of the non-upgraded 5%) mined an invalid block--as was an expected occurrence. Unfortunately, it turned out that roughly half the network hash rate was mining without fully validating blocks (called SPV mining), and built new blocks on top of that invalid block.
Note that the roughly 50% of the network that was SPV mining had explicitly indicated that they would enforce the BIP66 rules. By not doing so, several large miners have lost over $50,000 dollars worth of mining income so far.
The bolded points:
95% of the network was claiming to be mining v3 blocks,
50% of the network was SPV mining,
but only 50% of the SPV miners were claiming to be mining v3 blocks?
Doesn't that mean that 50% of the SPV miners, and so 25% of the network wasn't claiming to be mining v3 blocks?
If so, how did we reach the 95% trigger?