If you are interdisciplinary in the sense that you would call yourself a "Theoretical __________", "Quantitative ________", "Computational __________", or an "Applied Mathematician", then if you rank journals by impact factor you will artificially be biased against very good theoretical journals.
As a general rule of thumb (with several exceptions of course) the more theoretical the journal the lower the impact factor. The reason for this is because theoretical work is often more self contained. You cite a book or a few papers for the mathematical techniques or theorems you use, and not much else. Discussions and intros are short, the derivations are the crux of the paper, and the culture is that it is up to the reader to evaluate the importance of the results. It is not uncommon for a math paper to have less than 10 references.
For experimental papers you have to cite a lot of papers just to argue that your results are important. In addition, there are often many similar data sets that your result might shine light on or that contradict or confirm your result. In such cases you'd have to cite them all. Intros and discussions are often long and in many cases are viewed as the most important parts of the paper. It is not uncommon for an experimental biology paper to have 50-100 references.
In my field a good theory journal would have an impact factor above 2 and good applications journal would have an impact factor above 5. You have different audiences reading each. The reason I bring this up is that you mention as your example a general physics journal that accepts a lot of theory papers (i.e. papers with proofs) and one that seems to be more experimentally driven (and as an aside in a very hot field right now). The point is impact factors can be useful in comparing papers within a subfield but are not so good at comparing papers across fields.
If my result is of the form:
- Here is this theoretical result and it explains data from experiments or what we observe in the real world...
- Here is this theoretical result and it shows that all these experiments people are doing are missing the point. We should be doing these experiments instead...
- Here is this theoretical result and it means this for how we should be building ...
Then I want to publish in the applications or general science journal to reach a wider non-theoretical audience because the whole point of the result is its implications
If my result is of the form:
- We built this mathematical model which can be used to describe real world system; it has some really interesting behavior, but its unclear if this behavior is relevant to the system the model can be used to study
- We studied a mathematical model and proved X Y and Z. People have been using this model for a while and numerically have showed that it does a "good job" at ... we prove it actually does do a good job at ....
Then I would publish in a theory journal because the people you want to reach is other theorists who are comfortable with math. You want to get down into the gritty details.
Your philosophy may be different, but the point is that it is about reaching your target audience not the impact factor.