I think, though I'm not certain, that Wikipedia is in fact wrong in this situation, probably for the reason you just mentioned.
From this article, A square peg in a cone-shaped hole: The Samos E-5 recoverable satellite (part 2):
Herther had determined that a launch vehicle’s inertial measurement system’s gyroscopes would drift during the powered flight to orbit and therefore needed to be augmented with horizon sensors that told the guidance system what direction was up and down. The Agena therefore had horizon sensors as a direct result of his MIT research.
(note that although the article is nominally about the E-5 project, a separate satellite design, it used the same Agena upper stage that CORONA did)
I believe that only horizon sensors were used to feed the attitude control hardware. The exact orientation of the satellite was determined by looking at stars, but after the fact. The index cameras were side-looking devices which took additional photographs that were added to the film roll before and after the main camera fired.
Here's an example bit of output (thumbnail size only, alas) from The First Remote Sensing Satellite: Corona (Visible Imagery)

Note the horizon images at either end of the first and third strips. These would presumably have been used by the image interpreters to help precisely reconstruct the terrain being viewed.
edit and found this in Intelligence Revolution 1960: Retrieving the CORONA imagery that helped win the Cold War:
The guidance system consisted of three-axis gyros and Infrared (IR) sensors. These two IR sensors looked at the first space break point. Nitrogen cold gas jets acted as miniature rocket engines to drive the needed stabilization and pointing motion of the spacecraft and push the spacecraft as needed in no-gravity space.