We don't know which use case is being measured, but let's not kid ourselves.. the dominant use case is confirming one's pickup location on the map. Given the dominant use case, the map's reduced legibility = a worse map. (Map labels are used to confirm surroundings, and network isn't as important because riders aren't navigating.)
Also, comparing tweets to app store ratings isn't apples to apples. Everyone using uber's app has an app store account and can leave ratings. But not everyone who uses the app has a twitter account. (And ~40-50% of twitter accounts don't tweet.) So twitter results are skewed.
App store rating is the more meaningful metric.