Different CPU's will give different results. If you're online and offline will give a different result and MT makes up data as it goes along if there isn't any.
The back testing is useless.
You need to be on a proper language like JAVA and good data.
Higher time frames might be better, but I have never matched a single days trading to a back test for the same day. Try it, trade for a day or 5 then run a back test on those same 5 days and see if they are even close.
I tested my EA for years back using only M1 open tick data and then run it on real account for two months and the results were pretty much the same we expected. We had about 5 trades a day, TP somewhere between 10-20 pips. Market orders only. I used Forex Tester 2 for backtests because MT4 tester is so slow when you have more trades open (as we had). FT2 is about 10-100x faster depending how many open trades you have. But it really depends on your strategy how to test it to get reliable results. Sometimes it is favourable to just use bar close/open to do calculations because it filters out some spikes in tick data. And this speeds testing a lot!
I have found backtesting on MT4 to be a useful guide on live performance, but you will get differences on live and backtested results, due to broker slippage, down time and other factors. And yes, backtesting on the bigger time frames gives more accurate results compared with live results.
Backtesing proves several important points. One of the most important would be that humans fail to be consistent. We may have a system which works great for us in back test. Yet, in live our results never translate. In almost 100% of the cases it is because we gain bais to a certain position. Which in reality abandons the system which we are back testing. Many of us use 'rules' just as we input them into indicators or EA's [If this happens, then do this] Yet in live it is very rare to do the same. All in all, to make great amount of money in forex. You must trade like a robot. That is with no BIAS, or emotion towards any direction.
Backtest with real tick data then....real tick data is just the same as a live market. The velocity of the price, the exact moment of price surging up and down and the the way the tick behaves are all recorded in a real tick data. Whomever said otherwise doesn't know what they are talking about/have limited knowledge.