...and output at about 0.3
I assume you meant "
-0.3". For MP3, I'd actually lower that limit a little, as the filters used in the encoding process tend to create high-frequency peaks that can easily exceed 0dB. Technically, to avoid intersample overs entirely you'd need to set the limit to -3dB, but that's a little too quiet for the genre. Try around -0.5 or -0.6.
One problem you may be dealing with is that the two-track mix you're using is already mastered to a high average level, and you're adding a vocal on top of an already-mastered music track. That's going to tend to result in a flat-sounding instrumental, so you need to use as light a hand as possible in order to avoid doing damage.
I've never tried what you're doing, but if I did I would "master" the vocal and instrumental components on separate busses and then combine them. That way, I could do anything I want to the vocals, including smashing them flat with a limiter, without hurting the backing track.