Speaker frequency response is measure under standard anechoic conditions (or, at least, it shoudl be) using a pure sine wave input. The frequency of the sine wave, for a fixed amplitude (and thus power delivery / consumption) is changed from near zero to 20, 30 or even somethines 50 kHz, depending on the manufacturer and how diligent that manufactuer is in testing tis speakers.
What you should be looking for is a relatively flat line frequency response from 20 Hz to 20 kHz, which are regarded as the "normal" frequency range of the human ear, although we all tend to loose higher frequency perception as we get older. Any sudden spikes or troughs will indicate a harmonic problem, that is, the speaker cone(s) or something in the circuit is operating at or close to its natural frenquency of vibration. A speaker cone can have all sorts of different natural frequencies, depending on the mode shapes of the cone, ranging from the simple in-out movement of the cone to really really complicated ones with two, four, six, etc. different hot spots of vibration.
The linkey below is to an MEng dissertation, where the student was using finite element analysis to explore these mode shapes (quick explanation, the eigenvalues of a finite element analysis stiffness matrix correspond to the natural frequencies of the speaker cone, and from these one can calculate the mode shapes - the shapes the pseakers would look like if instantaneously frozen in time - look like). It's all highly testicle, but take it from a one time FEA engineer, it's true.
Linkey:
http://www.ewp.rpi.edu/hartford/~ernesto/SPR/Miller-FinalReport.pdf
The electronics in the crossovers can also have natural frequencies that can cause spikes or drop-offs. One of the eraly arguments for bi-amping / bi-wiring was that badly designed crossovers could / would / did leave gaps where the signal was in the range where it was too hig for the LF side and too low for the HF side, so feeding the full signal to both woofers and tweeters avoided such drop-offs. Mostly, good modern speakers and their crossovers don't suffer from that so much these days, but some speakers still do sound better bi-wired or bi-amped.
Any drop more than 3 decibles (dB) represents a halving of sound output, so you shouldn't be looking at speakers with drop-offs more than this in the 20 to 50 kHz range or above, say, 15 to 18 kHz and above.
Now. All of that is about how much sound is produced, not about what that sound, erm, sounds like. The tests are done with a regular, fixed, sine wave. They take no account of a moving frequency with sudden thump-a-thumps from a bass drum or the huge dynamic range of, say, a virtuoso rock guitarist's solo. That's when things like the dynamicity or sensitivity of the speaker come into play. Moreover, the speaker still isn't being fed a pure sine wave, rather a series of overlapping waves, that could confuse the speaker. Also, how much power is being fed to the speakers has its effect. That's usually measured as sensitivity also. I once had a pair of Mordaunt-Short MS15 speakers which sounded as flat and dull as you can imagine, until you cranked up the volume to neighbour annoying levels. Then the woofers started not only to vibrate, but to visibly move in and out on their rubber diaphragms. At that point the speakers came alive.
Bottom line: maths is great, but your ears are your master. And you need an amplifier with enough "oomph" to drive your speakers properly.