There may be a good reason for this
As India's detection systems got better, we started to detect and report more cases earlier resulting in apparently slowing recovery rates.
Example, let us say, On 5th April, we detected someone on the 5th day of his illness, and let us say recovery cycle was 15 days (as an example), this person would show probably recover by the 15th. But let us say on 6th April, we detected a person who was ill since 1 day, this person will recover only on 20th. Which is also why in the chart above recovery rates started at above 100%. When this % stabilises, it tells us we are doing a very good job of detection.
The assumption here is of consistent recovery periods, consistent treatment quality with increase in case loads and so on.
The only other reason recovery rates could drop is higher rates of deaths. But this is not showing in the data which means we are hiding deaths data (Highly Unlikely).