Ok this is an explanation taken from another website:
This is flash channels in play.
More capacity, more flash chips, more (parallel) channels the phone has.
Which translate to faster R/W, same behavior from SSD.
Usually max speed is hit when all the available channels are used (8 channels maybe?).
And to give good user experience, more mainstream capacity will be picked to utilize all the channels, in this case 128GB.
This translates to 128GB/8 channels (I assume) = 16GB/die
On 32GB variant, it could be using 8GB/die or 16GB/die flash, which mean only using 4 channels or 2 channels. Theoretically half or even quarter speed of the 128GB capacity variant.
For 256GB variant, it can use 8 channels * 32GB/die flash, of total 256GB capacity with same speed. This way manufacturer don't have to produce different flash controller to handle different capacity, changing the flash size will do.