If you use the price of the current tick in a TMF, you must indeed backtest with T1 data. But if you use only the close of the current bar, M1 data is sufficient - at least I would not know a reason of a "discrepancy".