From what I've gathered here and there, people that are a bit serious about backtesting use softwares such as WealthLab, Amibroker or MetaStocks, NOT mt3 or mt4.

Maybe, mt4 backtesting can be about ok IF you don't use any intrabar data. For example, if you work with daily bars, you should use only Open and Close values of daily bars (and not even High and Low if you set StopLosses or Take Profits). Nothing else.
The reason is, despite what their "modelling quality" might say, the only mt4 backtesting method that uses REAL data is "Open prices only".
"Every tick" and "Control Points" methods are nice fairy tales. Personnally, I won't trade money based on fairy tale data backtesting.
I'm still learning mt4, but in my experience, I wouldn't use mt4 backtesting in any other way. I'm not even 100% sure the way I suggested is 100% reliable. If I want to do good backtesting, I will get a backtesting software.

About mt4 vs mt3 for trading live: from what I've gathered here and there ;-) mt4 can't be trusted. Sometimes, orders that mt4 says were not executed, are in fact executed!! This was reported by brokers. If this is true, then brokers offering mt4 live trading aren't serious. Or, they made sure mt4 bugs didn't happen on their platform, it's possible too.
On the other hand, "I ve gathered here and there" lol that trading live with mt3 is possible.