If you know what the scales are and you know that you're interested in measuring those constructs, then you should stick with the measures you have, and not use factor analysis.
Here's an example: I construct a scale to use to assess the quality of a used car. I have some items about the tires (are they worn, are they cracked?), some items about the interior (Is it clean? Does it have holes in the seats?), some items about the engine, etc. If factor analysis tells me that I new tires and worn seats are associated and comprise a factor, I don't care. I'm going to keep my scales as they are.
If you want to know about the structure (or dimensionality) of the scale, then you should use factor analysis. But factor analysis never (in my experience) gives the structure that you expect. One reason is that factor analysis is an exploratory technique. There are an infinite number of solutions to the factor analysis of your data, and they're all equally good, as far as the underlying mathematics is concerned. What factor analysis does is tries to find one of those solutions that optimises certain criteria (and if you use, for example, a different rotation algorithm, you'll get a different result).
Usually you use some sort of structure (as you have) to make sure that you've represented the range of the construct of interest. For example, if I'm developing a math test for kids, I'll say that math includes: addition, subtraction, multiplication, division, fractions. And I'll make sure that my test contains items measuring each of those - I don't necessarily expect those to emerge as the factors.
So (probably) continue with the factors you have, but try to interpret them. Why did they emerge?
(Also, you say factor analysis, and then you say components. Are you doing principal components analysis or factor analysis? They're different things, with different goals.)
