That's because for MCFI to work it needs minimum of two frames to interpolate the third and the forth frame, which causes lag.
For an example, the latency between frame A and frame B of 24p content is ~42ms (1000/24). So 42ms + processing + insertion can mount to increased lag. So the audio must be compensated, which is difficult but not impossible.
To maintain correct phase and interpolation accuracy, it is better to buffer 5-10 frames in advance (at least according to a Samsung engineer, which does make sense IMO). Since Splash Player has direct access to the content, it shouldn't be too difficult as it can buffer the frames in advance (but may cause problem when seeking). The player also has access to vital data embedded within the video stream, which can be used improve interpolation accuracy.
HDTVs however don't have access to such data and have to wait for the information to populate, which is why it's not recommended for gaming
.
PS: I was referring to ReClock because video stutter witnessed in software based players is caused when the output is synced to the audio clock instead of the GPU clock. Random stutter can easily confuse MCFI as it's very sensitive to frame drops and uneven frame-rate. It needs stable and even frame rate to accurately interpolate new frames. Disturbing that flow can cause jumps in motion and artifacts. It would be helpful if the MCFI algorithm is adaptive and has certain intelligence, but we are not there yet.
Below is a quote that explains the function of ReClock and why it's essential:
"The purpose of ReClock is to definitely get rid of jerky playback of AVI and MPEG material on a PC (or a HTPC driving a TV, a flat panel, or a video-projector). It's a DirectShow filter which is loaded in place of the default directsound audio renderer.
It provides a new reference clock that is locked to the video card hardware clock, in order to ensure that frames are played at the exact speed of what is expected by the video card vertical sync. It also provides a frame rate adaptator for media files that do not match a multiple of the video card refresh rate (ex: playback of 23,976fps IVTC NTSC on a PAL TV). The combination of the two will give you the true experience of smooth playback with your PC.
Finally it is an audio renderer with hardware or software rate adaptation in real-time, multi-channel audio, audio timestretching (pal speedup compensation) and dynamic range compression capabilities.