A company deciding whether to fund a drug project is not asking whether the science is interesting. It is asking whether the expected return justifies the risk. Expected return is the value of a success multiplied by the probability of reaching it. Risk is the chance of losing the money already committed, and the size of that loss.
A project with a high chance of success and a large market clears the filter easily. A project with a low chance of success needs either a much larger payoff or a much cheaper path to survive the same filter. This is why two scientifically similar candidates can receive opposite decisions: they differ in probability, in cost to the next decision point, and in how much the market would pay if they worked.
The filter is applied repeatedly, not once. At each stage the company can stop, continue, or sell the project, and the decision is remade with whatever new evidence has arrived. A candidate that looked worth funding at the start can be dropped later if the evidence lowers its probability of success or raises its remaining cost.