Possible Errors With IAT

After talking about IAT's in class and personally taking a test, I was somewhat skeptical as to how efficient these tests actually are in showing people's biases. My results showed little variation in my preferences, and I just wondered whether or not there were possible errors with how these tests are administered. I came across ten common ways that the IAT is used improperly, found by Anthony Greenwald at the University of Washington:

1. Use of non-categories (unrelated words, nonsense words) as
presumably neutral categories in the IAT
2. Confounding category contrasts with positive-negative valence
3. Using stimulus items that permit alternate interpretations of
category contrasts
4. Treating subsets of IAT trials as measures of distinct
associations
5. Use of millisecond-unit IAT-effect measures (known to contain a
cognitive skill artifact)
6. Having subjects practice the attribute contrast before the
target concept contrast
7. Randomizing the series target and attribute items rather than
alternating them
8. Not counterbalancing order of administration of multiple IATs
when comparing magnitudes of these IAT effects
9. Making target-concept items indistinguishable in font from
attribute-concept items
10. Discarding error trials prior to data analysis

Do you think these are accurate errors with the IAT, or do you think that it is an efficient measure to determine how people discriminate others? I think a couple of these errors could be viably supported in most of the IAT's; however, I feel that the IAT's on race and gender are fairly effective tests.