Maybe the default random seed is not very good (the random generator is not good), and you can use another one, like \AMCrandomseed{1237893}?

Yeah, looks like it helped indeed.
An example:

Before:

After:

Doing this test, however, I found that maybe I wasn't completly fair. I didn't found copies sorted the exact same way than my source file as I said (I guess I was looking the catalog hehe), but I did found some homogeneous answer keys like the example above. On this source file, every correct answers were the first one.

On this another one, however, correct answers in the source file are sorted as A, A, A, D. The questions were shuffled, but the choices, not so much.

After changing the random seed as you suggested, it looks like it:

Not a big change, but I guess it's a bad example, as it's very small.

Changing the seed for 1587893 produced:

Gregório R. de Miranda wrote:

Yeah, looks like it helped indeed.
An example:

Before:

After:

Doing this test, however, I found that maybe I wasn't completly fair. I didn't found copies sorted the exact same way than my source file as I said (I guess I was looking the catalog hehe), but I did found some homogeneous answer keys like the example above. On this source file, every correct answers were the first one.

On this another one, however, correct answers in the source file are sorted as A, A, A, D. The questions were shuffled, but the choices, not so much.

After changing the random seed as you suggested, it looks like it:

Not a big change, but I guess it's a bad example, as it's very small.

Changing the seed for 1587893 produced:

\setdefaultgroupmode is set as {withoutreplacement}