A loss leader isn't very different than running in ad, in that you are simply paying for a desired benefit in lost revenue rather than in cash.
So, if you determine that a new client is worth paying $1000 to acquire, you might base an ad campaign based on the idea that spending $25k in advertising would result in at least 25 new clients which would be worthwhile.
You can take the identical mentality when offering products/services at a loss. Say you offered a free service/product that cost you $25k in expenses (i.e. the actual amount of money SPENT to provide those products/services, not the actual value of those services) but you wound up with 35 new customers. That could actually be a much better investment into new business as the example in the previous paragraph.
The trick is to find a product/service that you can offer a low/loss rate and still expect the desired sale or new client acquisition to occur. It's certainly a popular and effective way to promote business when done correctly.
I didn't understand your question 'what percent of a loss' is OK. It's not really about how much the discount is - it's more about the actual dollar amount that you are spending to make those loss leader sales happen. The percent doesn't really matter. You want to know how much money you are spending in promotions so that you can decide if it's worthwhile.