That's assuming the same algorithm is used throughout.

You can use one algorithm to encrypt/decrypt the original content. But you can use a different algorithm (with a different key) that would output a different output.

The secondary algorithm would be one that given a some text (ciphertext from the original encryption) along with the desired output, would return a suitable key. The most basic example to prove the point would be XOR.

That probably won't work: you need the combination of the algorithm + the key to get your data out. Of the 4 possible combinations, only 2 will yield valid data.

If you give a password, the cops will know what algorithm they must use (2 trials at most). Even if they don't know before hand which algorithm points to the real data, they can notice that it doesn't use all data.

With your method, you can at best cast doubt: is the data not extracted real data encrypted differently (algorithm or key), or random data that the software insert by default to give everyone plausible deniability?