Both choosing which issue to talk about and choosing which target to attack are issues of streamlining; I agree that there are things to gain by allowing them, but I think the slowed pacing outweighs the benefits. I'll explain the detailed reasoning for each of these issues as they currently stand.
At the moment, your chance of success when talking about the issues is altered according to the speaker, target, and issue; as Nenjin observed, a hippie lecturing a judge on law issues will be significantly less effective than vice versa. It doesn't, however, necessarily pick the best issue. I don't think it's very important that it does, to be honest. The bias isn't a gameplay mechanic, it's just a flavor tweaker. It helps to ensure scientists don't roll poorly on science issues so often, and a corporate executive is more likely to have a partisan retort about taxes than a biker. On the flip side, prompting the player to pick one of twenty or more virtually identical options is not usually a very fun game design motif, and I feel that speculating about what will work best for a given character may be fun, but it's only a little bit fun. Most of the flavor benefits from having that mechanic that are already realized without making the player choose an issue; you get all the fun of seeing your character talk about all sorts of issues and hearing a wide variety of responses when the computer picks an issue randomly. Plus, you're likely to see more of the content that way, with less repetition, since the computer's RNG is likely to spread its choices out much more than you are.
The reason combat doesn't allow targeting is that it would slow combat down and make it more complicated. Right now, the computer auto-targets, spreading attacks out a bit and preferring (but not guaranteeing) attacks against armed and dangerous targets, turning fighting into a one button affair that doesn't require a lot of thought, and emphasizing building characters rather than your tactics in battle. It can be irritating if your characters insist on shooting at the unarmed teacher instead of the death squads, which is why the computer has a subtle bias against doing that. For example, with four healthy cops, one crippled cop, and three nobodies throwing punches, each healthy cop will be shot at 20% of the time, and the crippled cop and nobodies will only be selected 5% of the time each. You'll still see some bad choices, but a few bad targeting decisions are an intentional part of the chaos. After all, nobody ever praised your rag-tag revolutionaries for their discipline and common sense.