Comment by Dove

In fact, the collateral fatalities from drone attacks are one of the highest at 98%, according to some estimates [1].

I think you're confused. The relevant paragraph in your link says the collateral damage -- both civilians and unknowns -- is 2% at the moment.

    The civilian casualty rate has been dropping sharply 
    since 2008. The number of civilians, plus "unknowns," 
    those individuals whose precise status could not be 
    determined from media reports, reported killed by drones 
    in Pakistan during Obama's tenure in office were 11% of 
    fatalities. So far in 2012 it is close to 2%. Under 
    President Bush it was 33%.
    
    Conversely, the percentage of militants killed has been  
    rising over the life of the drone program. The number of
    militants reported killed by drone strikes is 89% of the    
    fatalities under Obama compared to 67% under Bush.

I think that's fantastic news. A collateral damage rate of 2%, including unknowns "whose status could not be determined from media reports" is downright unprecedented in warfare.

The statistics, taken together, say they're using drones more in regular combat than they used to, and getting better at confirming that the kills are accurate. That's great news, isn't it?

There is a later paragraph that claims only 2% of drone strikes kill leaders. Maybe that's what you're thinking of?

    Since it began in 2004, the drone campaign has killed 49 
    militant leaders whose deaths have been confirmed by at 
    least two credible news sources. While this represents a 
    significant blow to the militant chain of command, these 
    49 deaths account for only 2% of all drone-related 
    fatalities.

I don't see how that's relevant, though. Is it somehow illegitimate to use drones on regular soldiers if it's tactically advantageous?