As "both" of your ranges are contingous, I think you can get away by testing whether the start and end points are "close enough" to a start/end point of a range. I would look at the "inversion lists" that the Unicode consortium proposes as an efficient way to implement (tests against) Unicode character ranges.
Some features that inversion lists provide, like fast negation, are not useful to you, so maybe you can get by with even less than what they provide. Maybe even just storing pairs (offset, length) for all your ranges and then doing a binary search for the range at hand already is enough. Maybe there even is a better data structure that allows you to simulatneously track start and end of your range and find the intervals they fall into, but I'm not aware of that.