The first Block I version of the capsule used three bags of the same diameter (43 inch). The later Block II used one smaller bag (34 inch). The volume of the small bag was 49.43 % of the larger ones. So they wanted to try a smaller bag with half the volume and calculated a 34 inch diameter bag.
The Apollo Command Module had two stable swimming positions.
Stable I was the desired position (apex up, heatshield down) with the crew hatch above the water surface. Stable II should be uprighted by the inflated bags to Stable I.
The Block I version vehicle weight upon splashdown into the ocean was approximately 9,000 lbm. Block II was much heavier, about 11,200 lbm and the center of gravity was shifted.
Tests were done using models scaled 1:5 and 1:10 and full-scale tests with boilerplates 1:1. Using finite elements computer models was impossible in the sixties. Uprighting should work too when only 2 of three bags were inflated. But a problem was found when one of the Y bags was not inflated.
However, a Y/Z bag combination resulted in a roll of the CM about its
X axis to a new stable position where uprighting did not occur.
Uprighting of the Y/Z two bag combination was improved by the smaller Z bag.
Source: APOLLO EXPERIENCE REPORT COMMAND MODULE UPRIGHTING SYSTEM
Crew relocations within the capsule were studied to be used in case of a single bag failure.
The recovery diver team was trained to upright the capsule in case of a dual or triple bag failure. The emergency CM uprighting was trained in the ocean using boilerplates. The uprighting sling assembly was fixed at the recovery sling of the capsule by the divers and a helicopter was used to pull the capsule in the upright position.
Source: DOD APOLLO RECOVERY OPERATIONAL PROCEDURES MANUAL
So the size of one bag was decreased to optimise uprighting performance in case of a bag failure.
But the Navy was prepared to use a plan B when all bags failed.
In 2011 a paper Simulation of the Apollo Command Module Uprighting System using LS-DYNA about computer models for the Apollo uprighting system was published.
The commercially available finite element analysis code LS-DYNA
was used to generate a simulation that accurately models the
command module vehicle uprighting dynamics.
Additionally, the buoyant characteristics of the command module
are also independently verified using the Airborne Systems
proprietary statically determinate buoyancy solving code,
The uprighting system of the Orion capsules was studied using 1:4 scale models, see Dynamic Characterization of the Crew Module Uprighting System for NASA’s Orion Crew Module.