Etesla,
I looked at the data sheet for the MC12093, that is hopefully representative of the series. Reading between the lines I think the low frequency limit of 100 Mhz is a result of the extended rise and fall times if a sine wave is used. I expect that if the sine wave is squared up by an amplifier and voltage clamp, the prescaler will operate correctly at lower frequencies. These series of chips are implemented in ECL that, unlike TTL or CMOS, are not saturating and don't really have much voltage gain. Any input signal that has slow transition times relative to the switching times of the circuit, as does a sine wave, will be susceptible to induced noise or to misinterpretation by different elements in the chip. It's too bad the data sheet doesn't seem to specify maximum input rise and fall times, as do most, if not all other digital chips.
BTW, do you have a particular divider design you'd like to implement?