Burst mode does not mean the entire file is transferred from disk to main memory all at once. This would be impractical. What burst mode means is that, once the DMA is ready with some data from the I/O device and is granted bus request from the CPU; it will transfer the block of data which it has obtained from the I/O to MM. The size of data transferred in one go, is usually defined by the word count register.
As given in the question, the WCR = 16bits. So, in DMA burst mode we can transfer 64KB of data before DMA releases the bus. This however, still leaves the CPU inactive for that entire duration of transfer (which is quite long compared to cycle stealing).
In cycle stealing mode, one byte ( and not the entire block of data, represented by the word count register) is transferred before the DMA gives back the control of bus back to the CPU.