@ Mithun,

So many researchers are faced with the same issue.

Please go through these parallel rooms. You will find some directions and nice tips with detailed explanations on how to overcome your concern/query.

Raoof

https://www.researchgate.net/post/Filling_of_Rainfall_data_gaps

https://www.researchgate.net/post/What_is_the_common_framework_for_filling_missing_data

https://www.researchgate.net/post/What_is_the_most_effective_way_to_fill_the_missing_hydrological_data

https://www.researchgate.net/post/How_to_calculate_missing_daily_precipitation_data

https://www.researchgate.net/post/How_should_we_deal_with_the_missing_data

https://www.researchgate.net/post/Missing_data-how_to_fill_them

Fill gaps at your peril!  Sort of depends on what you are trying to do, but I've yet to find a fancy model or algorithm that can handle geophysical events that are highly episodic (like rainfall).  Ask yourself "What are the consequences of getting it wrong?"  If you miss an extreme event, how much does it affect your conclusions.  For some applications this can be a big problem - so don't try and fill gaps with any technique other than real data.  Examples might be design criteria (which are very dependent on extremes), model fits (which can be seriously biased if you get extremes wrong), and reporting or using summary statistics (which sometime use extreme values for serious decision making).

Better to try and find a way to get where you want without manufacturing data - and here you just need to be smart, and it depends very much on the end result you need. (Another useful technique is to conduct a sensitivity analysis - if you have the wherewithal and time,  Re-run your case with different data in the gaps - however you fill it - and see how much it affects your conclusions).

Yes, 

It is time consuming, tedious, computationally costly but still prone to error.

Depending on the accuracy you look for the data gap, you can opt for downscaling several times by refining the resolution in each nesting with physical based weather/climate models. 

But, as a matter of fact, you are looking for the gaps. But, you remember that data you have also have errors.

In a nutshell, it is about what you are studying is really a matter. If you can clearly see your target, you can understand if you need to go for the gaps or not.

Hope it helps !