Valuation isn't really a science. One can always find ways to justify a valuation. Let me give an example taking up DCF. Try varying the growth rate in any DCF model projecting 10 years of cash flow. To my mind, unless there is a good amount of certainty of future cash flows, DCF may not work. A road toll project is where DCF works fine, as the CAPEX and future cashflows can be reasonably defined and are sometimes backed by a master contract.
There is a lot of literature available citing how investment banks do valuation based on which side they are on - buy or sell.
Another example would be acquisition of whatsapp. According to me there is absolutely no way I can come to that valuation (of USD 16 Billion i think), specially when it is free for foreseeable future. Perhaps, facebook thinks it will be able to earn revenue in the long term.
In your example, perhaps Zomato is overvalued and JD is undervalued. Perhaps Zomato is not able to monetise its offerings the way investors thought it would. One can project future cashflows, but when it comes to actually realising them, there can be a lot of difference and more so in case of internet companies with novel offerings with very few precedents.
Final thought, a lot of internet companies are built just to be sold. That is the reason investors are interested in keeping the value continuously high. In case of valuation, it isnt math and logic all the times
Thanks for brining up such an interesting topic.