I use LightCycler 480 (and also 2.0 but this one is not prepared for HRM analyses). Unfortunately I've only compared this machine with Eppendorf's Mastercycler realplex, not with ABI-7500 or RotorGene 6000, so not much help there.
Hovewer I can write about my experience with systems I used.
I very much like the software provided with LightCycler 480, considered as one of the best in the market. The results are visualised better than with Ep realplex, but that's my subjective opinion. With quantification analyses the software gives you standard deviations, mean/median values, standard error values etc. Everything is dynamic, intuitive (I seldom used manual, really) and, what's very important - the service is there for you (in our case at least).
I use the lowest volumes per reaction possible and everything works just fine
For this system other kits can be used, not necessarily made by Roche, but Roche's always give the best results.
It is 7-channel system, which may be of some importance, and the heating-cooling system (block) gives you high temperature homogeneity. Of course it is prepared for HRM.
Where is the catch? The system is regarded as rather expensive. And it's not the machine itself, but kits, which are a bit more expensive than others - in the long run that matters more. The good thing is that kits for LC 480 are twice as cheap as for LC 2.0, so they are improving