Here is the basic reason why the selection of fillers is so important:  Eyewitnesses tend to select the person from a lineup who best fits their memory of the perpetrator relative to the other members of the lineup.  As long as the actual perpetrator is in the lineup, this "relative judgment strategy" works quite well.  Unfortunately,  when the perpetrator is not in the lineup there remains nevertheless someone who better matches the witness's memory than do the others.  The purpose of the fillers is to try to make sure that the chances that an innocent suspect will be the "relative best choice" are 1/N, where N is the number of persons in the lineup.  Suppose, however, the suspect is the only one who actually fits the verbal description that the eyewitness had given of the perpetrator.  Such a situation virtually guarantees that the suspect will be the "relative best choice."

Consider the first lineup below.  In this Texas case from 1995, the perpetrator was described as a black male.  You might not be able to see this very well on some monitors, but there is only one black male here, the rest being hispanic.  Sad, but true.

Shame on these criminal investigators in Texas. 

Consider now an example of an excellent photo-lineup:
Richard Vulgamore, a special investigator with the Brooke County Prosecuting Attorney Office in Wellsburg, West Virginia, came across a problem when a critical eyewitness described a perpetrator as a black male with crossed eyes.  Other than the suspect's photo, which had crossed eyes, they had no photos of crossed-eyed black males.  Vulgamore used a computer and forensic art techniques to create crossed eyes in photos of black males that they had on file.  The result is pictured below.  Can you tell which person is the suspect?  The eyewitness was immediately able to identify the perpetrator.

Kudos to Vulgamore.