Somebody else may know better, but as far as I know there's yet no simple to use piece of software that allows what you need. Polyphonic pitch recogniton software exists with pretty sophiscated capabilities, but they cannot guess what part of the signal is of interest to you or simply distinguish a melody from accompaniment. In simple terms, the sw is going to try do identify clear harmonic components (the type of sound generated by pitched musical instruments) in the signal and detect the base picth (the so called "fundamental") for each harmonic component. Instruments in the same pitch range and the presence of other types of signal, like noise or percussion instruments may make this task extremely difficult, providing chaotic results like the one you show.
It's possible to "help" a sw tool by telling what parts of the audio signal you are looking for, but it's an interactive error and trial process and you have to know a bit about audio signal processing to understand how the sw works and how to provide it with useful information. HPCP for example is such a tool, but it's an engineering more than a musical tool and will not by itself provide you with musical notation.
Even if you can isolate the harmonica component of the tracks, an aditional difficulty is the characteristic expression playing of the harmonica (bends and draws with sub tonal inflexions) that will make pitch recognition more difficult, as pitch recognition sw looks for a stable pitch for a specific amount of time before declaring it a found "note".
Again, someone may have a better answer but I don't see an alternative to the good old method of going about it by ear.