When attempting to reduce the rate at which Steam dynamos consume Steam at first via KubeJS, and again by datapack, I found that my attempts to reduce the recipe's requirement to 500 mB instead dropped the steam intake per recipe much farther down to 100mB.
Test Datapack

This result was the same whether the recipe was modified by placing the full datapack in the saves folder, the contents of the datapack in KubeJS' data folder, (An automatically-loaded datapack for modpack development) or the recipe was altered via a KubeJS script.

This was all confirmed to take place on an instance with only Thermal Systeams, JEI, and their dependencies.
Minecraft version: 1.20.1
Thermal Systeams version: 1.7.1

What is the energy amount showing for that example? I use Thermal's recipe system, which performs normalization on recipes so that they show nice amounts of input. (So for your example the steam might show 100mb input but the energy will show 400, one fifth of the amount you set for 500mb of steam)

No, it's still 2000 FE.

Of course this could all be JEI misrepresenting the actual recipe, but it'd still be an issue.

I should also mention that I confirmed the actual recipe to consume only 100 mB per recipe, but on an an instance with a ton of other mods. Since I did that before switching to the instance with only Thermal Systeams and JEI, I had thought it might've been the result of a cross-mod interaction

Can you try a recipe with the same amounts in a normal thermal dynamo?

Of course.

From my tests with the datapack at the top of this issue with default configs:
1 coal generates 24k RF in a Stirling Dynamo (as expected from JEI's recipe)
1 coal generates 24 Buckets of steam in a Stirling Boiler (Despite the config supposedly halving that?)
24 Buckets of steam in a Steam Dyanmo produces 480k RF, exactly as JEI reports the recipe to be (100mB steam to 2k RF)