My intra-day system development took a setback due to confusing 5M data. I'm now testing a Tokyo open, range breakout system. The problem is that pinpointing historical opening times can be very confusing due to changes in daylight savings, AND the fact that brokers aren't always clear in how they set their server time.
It took a whole day to find what I think is a solution to this problem. I use Forex Tester 2 to backtest historical data. For actual data to backtest on, I've used a standard subscription with Forex Tester, which gives me 5M data from different brokers like Alpari, FXCM, Forex.com etc.
The best data for intra-day trading, IMO, is FXCM. Their server time is set to GMT, so there's no confusion over daylight savings (unlike other brokers, which may set their time to local time or the NY close). This allows you to accurately pinpoint the start and end of various market sessions. Plus their data is 5-decimal digits (or 3-decimal digits for JPY pairs).
As for finding actual market session times, there are threads by Clint on Babypips that contain this information. e.g. Forex Trading Sessions --- September 2012 - April 2013. You can use the search function on the forum to find times for other months. This saved me alot of headache!