Stating you can't blow or damage a speaker when running less then rms power is like stating you can run more power and never damage a speaker. If I have a amp rated at 100w rms and the speaker is rated at 99w I highly doubt manufacturers rate their rms rating at its true max..... Has to be some fudge factor/safety margin in their number.
There is, but what's important to understand is that the rating is "Average power over time". A speaker designed to handle 100w of power means that it will handle 100w continuously for an indefinite period of time without damage. As you start pushing power levels above that, you are adding more themal energy into the speaker than it can dissipate in a given period of time, the speaker will begin to heat up beyond the levels its designed and can fail. It's not going to happen immediately though.
If you were giving that speaker 200w of power in small bursts such that the power averaged out to 100w or less, the speaker
generally would be able to handle it no problem. For example, giving a speaker 200w of power for one second, and then 0w of power for another second will have an average of 100w of power over the 2 seconds. Something that the speaker should be able to dissipate thermally assuming something else catastrophic didn't happen.
Music is dynamic. Most music has a crest factor of at least 6db, which means that the average power is 1/4 or less of the peak power. If you had your 100w speaker on a 200w amp, the speaker is still only receiving an average of about 50w. This is a highly debated topic even on these forums but most people agree that doubling (or more) the rated power is not an issue. The one exception to this would be subwoofers on bass heavy music where the crest factor can be much much less.
As far as "underpowering", it's been discussed well here already, but maybe this makes it a little clearer.
a 100w amplifier can produce 100w of clean un-distorted sound. If you were to drive it DEEP into clipping, the amplifier will produce 200w of garbage (assuming the power supply in the amplifier is up to the task). P = Vavg^2 / R, and Vavg = 1/sqrt(2) * Vmax for a sinewave => P = 1/2 VMax^2 / R
As you drive the amplifier into clipping, the average power approaches Vmax. Worst case, the average IS Vmax. Pclip = VMax^2 / R
Pclip = 2 * P (absolute worst case scenario).
So your "underpowered" amplifier is suddenly an overpowered amplifier. Better yet, if you're NOT using any sort of passive filtering on your speakers (like you have an active setup), that clipped signal is now full of energy all over the spectrum, causing your speakers to play lots of frequencies they weren't designed for. This means the very real potential to reach the mechanical limits of your smaller drivers.