Sure, a voltage source in in series with a sufficiently large resistor (compared to the load) behaves like a current source (Thevenin/Norton). However, given the frequencies involved and the currents, I see problems.
To have a decent amplitude accuracy at 100 MHz, you would want your load impedance as seen from the generator to be close to 50 Ohms. This would require a transimpedance amplifier input impedance well below 5 Ohm, but would require voltages down to 500 uV. And obviously the capacitor in parallel will screw up your frequency response (depending on the value). So I doubt it would work particularly well.
Maybe if you calibrate your system over its frequency range. And regulation will not be amazing due to the low source impedance. Maybe coupling the signal through a transformer could improve matching and regulation?