There is no contradiction in two approaches.
Baum-Snow (2007) looks at the effect of the Interstate Highway construction on those MSAs which happened to lie on its way. The randomness comes from the fact that when the plan for the Interstate Highway project was considered, the aim was to connect distant areas. Through which MSAs the highway would go was more or less random (at least this is what the author claims). Thus, some MSAs received treatment by lying on the way whereas other didn't.
Donaldson (2012) uses planned railway routes that were NOT implemented for a placebo test. The concern which Donaldson addresses is that there could exist some unobservable economic factors (different economic development, agricultural income etc) affecting the decision about where to construct the railway. To overcome this, the key assumption he uses is that, on the stage of planning, the several proposed routes were equally desirable from the strategic point of few, but some of them were more costly to build, hence these routes were dropped at some stages of construction.
Now, if we saw an effect on income in areas where railroads were planned but never built compared to areas where railroads were not even planned, this would suggest that railways are probably not the reason to explain main findings in the paper, but rather the selection of more developed areas into treatment (receiving or at least planning to receive railroads) was in place. However, such placebo test shows that areas that were initially selected into construction of railway, but due to technical reasons didn't receive the railways - have no significant income effect. Hence, the selection is argued to be not an issue in Donaldson's identification.