Some comments on the 25-35Ohm range for the impedance of Yagis, which
has VE2PID from my website. Without further details this makes indeed
no sense to say this would be the best for Yagis.

1. You must differ between short Yagis and Longyagis.
2. You must know for what bandwidth the Yagis should be designed.
3. A given result you can reach with a nearly infinite number of
changed parameters.

For short Yagis and the bands 14-1,35MHz, 21-21,45MHz, reduced 10m-
Band from 28-28,8MHz, and for 144-146MHz a low impedance (10-15Ohm)
does not cover the entire band, but gives high gain. A high impedance
with 50Ohm gives away to much gain for a to high bandwidth. For
example a 3-Element-Yagi with 50Ohm direct feed cannot reach more than
5,5dBd. The reasons therefore are pointed out by Roy, W7EL in his
comments about element coupling and element currents.

For the enumerated bands the impedance of 25-35Ohms is indeed the best
for a good balance of gain, pattern and bandwidth. By adding a close
spaced D1 for 50 Ohms you can reach similiar results, but why doing
that? For more weight and additional mechanical problems?
It is better to match the impedance to 50Ohm than to add more elements
for rising the impedance!

For Longyagis the problems are more difficult. You must see the
radiator, D1 and D2 as a unit (radiation center) in a Yagi system. The
very close spaced D1 in a 50Ohm-feeding system acts like an "open-
sleeve-element" and has higher currents than the radiator. The
interaction between these elements can reduce the bandwidth
dramatically, because -j and +j of the impedance increase very fast if
you leave the center frequency.
It is interesting but a fact: You can replace the 50-Ohm-radiator and
a close spaced D1 in several Longyagi systems by one radiating element
with lower impedance and greater bandwidth.
Evolutionary algorithms for optimizing Longyagis find more lower
impedances than 50Ohm for a given gain, pattern and bandwidth. The
matching "by hand" with the close spaced D1 (as Ian, G3SEK writes) in
the finishing construction process is possible but no must.

73 de Martin, DK7ZB