You're in a foot-race between forward losses ("diode" Vf*If) and
reverse losses ("diode" low-Vgs / reverse Vgs leakage). With so
small a max voltage you can expect not-good forward losses.
With some modern technologies, GIDL phenomena can make a
negative Vgs raise the leakage floor as you go more negative
and this is worsened by the same short channel that you'd want
for best forward operation.
I'd approach this as a "sweet spot search" (optimization) with
as much realism regarding the true nature of source and load
as you can wrap around the "energy harvesting core". As one
point, with a discharged load you may find that the "antenna"
fails to make the purported 0.25V, even, when the load is
"seen" through the "rectifier". This can lead to startup issues.
You may need to do things like "UVLO" functions that detach
the load from the charge reservoir until adequately pumped up.
The trade between wide (for forward efficiency) or narrow (for
leak-back efficiency) depends a lot on load and source "box".
If you can (say) make a rung-up tank circuit to drive a higher
VT gate to better "on" and "off" positions, that can beat a fixed
diode bridge. However making this work at 2.4GHz would probably
add more commutation losses than you might stand to gain -
another thing that wants analysis, to say.