This type of anecdotal example is too little information to draw conclusions about relative strength of ED boost. For example, you mentioned that many times the admits are legacies or athletes, so it would be good to separate legacy/athlete vs unhooked… Assuming unhooked kids are admitted, you might review how many top stat unhooked kids applied ED vs RD and their relative admit rates.
In the national pools, these colleges do certainly admit kids in the RD round. At nearly all ED colleges, the vast majority admits come from the RD pool rather than the ED pool. Specific numbers you for the colleges you mentioned are below, as listed in the most recent CDS, assuming near 100% yield for ED. Note that the relative admit rate difference does not consider the higher rate of hooked kids applying ED than RD. The actual degree of boost for ED is probably significantly less than suggested by the relative difference in admit rates.
Northwestern ED – 1077 admitted with 27% admit rate
Northwester RD – 2345 admitted with 6% admit rate
Overall Admit Rate = 8%
Duke ED – 879 admitted with 22% admit rate
Duke RD – 2310 admitted with 7% admit rate
Overall Admit Rate = 9%
Vanderbilt ED – 852 admitted with 21% admit rate
Vanderbilt RD – 2446 admitted with 8% admit rate
Overall Admit Rate = 10%