These are different problems.
1. In firmware there is a max. feedrate for x and y. If this is set higher you can move faster. No problem just make sure you do not loose steps, otherwise reduce max. speed.
2. It moves with the speed you send in gcode. In host or server you can set speed for manual moves for example. If these were reduced before because original firmware had lower max. speeds set it might explain why you now see faster moves.
3. Slicer print time calculations are always wrong. They neglect things like max. feedrate and acceleration/decelleration so they are always only a coarse guess. Depending on infill pattern and shape they come closer or not. Repetier-Server simulates such effects if configured correctly and can achieve quite good time approximations, but even there you get some uncertainty from heating up and different firmware implementations.
