Compare apples to apples and oranges to oranges, the closest your quoted post provides to such a comparison is as follows:
15400tx (blocksize 3.5 mb) @ vs 15840tx (segwit w/ 2mb)
It's not accurate since 3.5 mb < 2mb segwit, but none of the other numbers line up at all.
But " I propose we work immediately towards the segwit 4MB block soft-fork" (Gmax)
https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2015-December/011865.htmlsegwit is 4mb.
that was when Gmaxwell thought the plan was 2mb blocklimit AND segwit.
in the same paragraph he said
If widely used this proposal gives a 2x capacity increase
which is where he got the 4mb number from. but more recently its suggested to be only 1.8x capacity... hense new numbers are 3.6mb(real data) for 2mb block limit,
also when using the 1.8x in combination with the average tx size brings out the 7200tx a block everyone is throwing around as segwits promise..
yet listening to Gmaxwell and Luke JR in the last month.. it looks to be 3600tx for 1.8mb(realdata) while sticking with just 1mb block limit