Because bias, and the word score boost attributed to bias, was equally distributed between participants who received the drug and participants who received the placebo, if the drug worked, the group that received the drug would still score higher than the group that didn't because their scores would combine the bias boost with an actual drug boost.
In other words, let's say the boost from bias, or poor design, or lying, is X. And let's say the boost from the drug is Y.
In the trial, X is evenly distributed between drug recipients and placebo recipients.
If the drug worked, the group that got the drug would score even higher, X + Y, because they would get a boost from the drug on top of the boost from bias. But that didn't happen. Everyone showed the same boost, the boost from poor trial design, bias, lying, however you describe it, with no advantage for participants who received the drug.
Therefore, it doesn't really boil down to your two options because they don't take into account the fact that no one knew which group they were in. Poor trial design boosted everyone, and what we're left with is a drug that failed to produce a signal on top of that.