R&D is not omniscientAnother explanation is that the developers of a collectible game often can't say themselves what will turn out to be good or not good. This is especially the case early in the game's history, when no one — including its creators — really understands what makes the game work. The contrast between the cards from the first Magic: The Gathering set that are considered powerful now, and those that were considered powerful then, is very instructive. And sometimes people just make mistakes. Magic: The Gathering provided the perfect example of this with the Urza's Saga block, which was created in order to slow down the game, but produced a Magic World Tour ("Combo Winter," Rome 1998) in which tournament matches were decided by the initial coin toss, because both players' decks could win on turn one. Sometimes, too, a player shows up at a tournament with a deck built around a card that everyone else thought was worthless, and proceeds to mop the floor with his rivals and send the price and reputation of his key card into the stratosphere. In Magic: The Gathering, Necropotence, from the Ice Age block, is the most famous of these. (And sometimes you have lead designers who enjoy seeing how much they can get away with, whose antics periodically break the game in manners that must be incredibly funny to those who are neither fiscally nor emotionally invested in it.)
Analysis / Junk Rare
For an in depth look at why this trope exists, at least in relation to Magic: The Gathering, and probably most other Collectible Card Games, please see this article. When it comes to collectible games, not all the cards can be equally good. If they were, it would continually up the power level of the game as well as remove the strategic aspect of creating a deck - one of the main draws of the genre. It's also not economical, either for the manufacturer or the players, for only rare cards to be good. In order to hook and maintain the audience, some of the cards must be less valuable or useful. Good cards are spread across all rarities as much as possible. Since the criteria for acceptance as a rare isn't just the card's usefulness, this will naturally lead to some rare cards being useless, or at least so specialized that they are difficult to use effectively.