@Joe can you rerun the test with the AIs on challenging? There is a night and day difference as to the AIs play and early game expansion.
"Hard" is actually a difficultly level that is harder than "challenging" (or were you wanting me to decrease the CPUs abilities?). I did bump it up to "RIDICULOUS" (highest) and ran the tests again on 4 new maps (2 trials per map):
HERE are the compiled averages from 8 "runs" across four maps (including "old" data from HARD as well for comparison)
Starting by OASIS/WILD GAME vs. FERTILE LAND*:
Population: 76.5 ** vs 122.8 (p-value = <0.05) (HARD)
Population: 70.4 vs 113.5 (p-value = <0.05) (RIDICULOUS)
Techs: 2.0 vs. 4.4 (p-value = <0.05) (HARD)
Techs: 9.4 vs 10.5 (not statistically significant (RIDICULOUS)
Power: 18.2 vs 23.9 (non-statistically significant trend) (HARD)
Power: 28.8 vs 37.1 (p-value = <0.05) (RIDICULOUS)
Settlements: 3.5 vs. 3.2 (not statistically significant) (HARD)
Settlements: 2.0 vs 2.5 (not statistically significant) (RIDICULOUS)
(1) Population difference holds up under both HARD and RIDICULOUS.
(2) Under RIDICULOUS a significant shift in the CPU priorities occurs to TECH upgrades, which seems to trump army building as evidenced by big jump in focus (2.0 -> 9.4 and 4.4 -> 10.5):
(3) Under "HARD" difference between starting positions didn't affect "POWER", but CPU shifts resources away from troops to build up tech under ridiculous.
(4) Shift to tech specialization also seen in the development of LESS settlements.
EDIT: Additional note -- there is a ton of variablity on AI growing population/power...regardless of whether there is FERTILE LAND or OASIS ~40% of the time by turn ~70 the AI only has population of "10" ?!!?!