I believe Arduino PWM signals are about 500 Hz? So well within the 1MHz bandwidth but a very narrow duty cycle (near 1% or 99%) could be missed.
5 million samples/second technically puts an upper limit of 2.5 MHz on the bandwidth. I'm guessing the analog-to-digital converter (whether they use an external one or the AVR's internal one) either has a lower bandwidth or they place a cheap RC filter prior to the A/D to prevent aliasing and that has a cutoff frequency of 1MHz so that there is at least some attenuation (~25dB) at 2.5 MHz.
All in all, you get what you pay for, but for teaching students about time-varying waveforms $49 is not a bad deal. And I'd much rather have a high-school student "accidentally drop" a $49 scope than a $4900 one 
The only drawback is that if someone doesn't understand about bandwidth and they look at a signal on the scope to draw conclusions from it, their conclusions could be way off if the signal they're looking at has high frequency content. A little knowledge (or insufficient bandwidth) is a dangerous thing.