Actually, if it is a CMOS/HCMOS output, it might be better to terminate
it into a voltage divider of, say, 270 ohms from the osc output to 56
or 62 or 68 ohms to ground. Many of the common clock oscillators are
not intended to directly drive a 50 ohm load. Then an output taken
across the resistor to ground will look like it's from a nominally 50
ohm source. You could use a larger voltage divider ratio to get the
output down further if desired. If the osc has square wave output,
that's likely OK for a mixer input, but if it's not 50% duty cycle, you
might benefit from cleaning it up a bit with a tuned circuit, for
example. And if the oscillator has a TTL output (rather than HCMOS),
you might benefit from returning the voltage divider to the osc power
supply, presumably 5V. Beware when calculating the power delivered
from the voltage divider to a (say) 50 ohm load; put that load in
parallel with the output resistor of the divider to calculate the net
division ratio. So the suggested resistors, e.g. with 68 ohm output R,
might give you about -8dBm, if the osc delivers 2.5V P-P into the
divider, yielding roughly .24V P-P output into a presumed 50 ohm load.
-- Also, the oscillator probably has DC on its output, and you might
benefit from a blocking capacitor. 1000pF would be adequate. And
beware that the osc might deliver noticably higher voltage into a
resistive load.

Perhaps the OP could provide a bit more detail...

Also, I'd ask if that -16dBm is accurate...that's pretty low even for
active mixers.

Finally, RFSim99 is a nice little free program for playing with linear
RF circuits, and includes an "RF calculator" which has a tab for signal
levels, converting among dBm, watts, volts-RMS, and volts-P-P for a
user-specified impedance level. (I wish that tab had "locks" on the
values like the resonance one does, so you could see what power level
you get when you load a fixed voltage with various resistances, but you
can always just copy-and-paste the voltage to "remember" it over a
resistance change.) I always appreciate that RFSim99 has a lot of
tools all in one place, and I don't have to remember a whole bunch of
different programs, each of some very limited scope.

Cheers,
Tom