In No Limit Hold 'em: Theory and Practice by David Sklansky and Ed Miller p106:
But every once in a while you should semi-bluff raise to perhaps $40 or so, and you should do so with 7c2d and not Qh8h.
Why is 7c2d a semi-bluff?
I think that they are using the term "semi-bluff" pretty loosely here. As I read it, I take it to be due to the fact that you still have a fair amount of equity against most of your opponents' calling range.
Often you actually have better equity with junk than with mediocre high-card hands, since 7♣2⋄ will probably not share any cards with your opponents, but they may well hold a Q, making Q♥8♥ a disaster when you hit a Q.
The 7♣2⋄ is a "semi-bluff" because it has a better possibility of being good when it hits post-flop. It still seems like a stretch to me to call it a "semi-bluff" but I believe that's the logic behind the statement.
Qh8h is a pretty good hand, two-handed. For this reason, it should be played in a straight-forward manner.
7-2 offsuit is recognized as a "trash" hand. If the flop doesn't "pair" either of you, it can't beat, say J-5 (a statistically average hand), whereas Qh-8h can. So you bluff with it, and hope that your opponent folds.
Two-handed, it's a semi-bluff, because if your opponent has, say A-K offsuit and doesn't pair (or better) on the board, and you do, you'll win. And if 8-5-2 of mixed suits shows up on the flop, no one would put you on a pair.
n.b. I'll likely edit this answer when more detailed info is given in the question. For now I'll keep it short.
From the original question and looking at the comments, it seems that the issue revolves around post-flop equity.
Q♥8♥ has more post-flop equity than 7♣2⋄.
Specifically, given the opponents likely wide ranges and your positional disadvantage, it would seem better to raise with a weaker hand utilizing your fold-equity to win the hand now, and better to call with a stronger hand to utilize greater post-flop pot-equity. The optimal frequencies for these plays is a subjective issue significantly dependant on table dynamics. You comment the table is tight, which is a good start.