Values are different in Shogi, where the drops and 6th-rank limited
promotion rules change all values. After I sought out and visited the Shogi
equivalent of the Nihon Ki-In (but decided not t play because of my limited
shogi experience -- the exercize of finding it on the map and actually
finding my way to a remote region of Tokyo was quite rewarding enough!), I
downloaded the supremely weak shogi master program from the home of the
underdogs, and played quite a bit; and my impression is that (1) having
more Pawns is good, but not specific number of Pawns is worth anything; (2)
fR == ffN; Gold == Silver == 2 of the previous; R == B == 2 of the previous
(although R versus B may often be decisive).

But in a 'normal' game, what's a Silver worth? My ancient researches seemed
to indicate that adding the forward move+capture of Wazir to something else
is worth nearly half as much as adding a whole Wazir; and that adding the
sideways moves is worth most of the remaining half-Wazir. For the Ferz, I
forget. It's written down somewhere but of course forwards is worth more
than half.

Thus, the basic estimate for the Silver General is roughly 3/4 of a Knight,
and the basic estimat for the Gold General is somewhat more than 3/4 of a
Knight.

If the simplest possible estimates of the values are taken, then your
Elephant may possibly be worth 3/4 N and your Great Elephant may possibly
be worth 1.5 N, which is roughly the value of a Rook.

According to the shock-troop theory, the combination of weak FfW which have
no jumping move and the strong fWfDWA which does have jumping, this
combination interferes with the harmonious development of the army; but
Philidor's shock-troop theory, although it contains some truth, is not the
final word. Morphy showed how one can cause great damage to the opponent by
developing the Rook-valued pieces, and his example must be kept in mind
when you are playing an army that includes HFD or Great Elephants as
R-valued pieces.

If the Great Elephant is Rookish in value, then it is logical that splicing
equine genes into the Greater Elephant must produce a Q-valued piece; using
this in the same army would be consistent but it would be overkill.

If my guesses about the elaphantine values are somewhat near correct, then
why does your playtesting seem to hint that the army is too weak?

In my experience, the unreliable values produced by my pseudoscientific
guesses are actually more reliable than playtesting. The most common
problem with playtesting is that if you don't know how to use a piece to
best advantage, it seems weaker than it really is. Learning to use every
possible piece is difficult and time-consuming.