The EMH is just a convenient tool for analysis, not really an attempt to describe real markets.
If you want to do any kind of analysis you have to make an assumption about how prices are set. Assuming they reflect availaible information is a reasonable choice. If everyone admitted "We have no clue what influences prices" then the whole field of finance is dead. Basic concepts such as alpha/beta, diversification, risk/reward and more all have their roots in EMH.
Note that anomalies like value or momentum investing can coexist with EMH, assuming that the additional return is offset by increased risk (ex: small-cap anomaly). Similar arguments can be made for momentum (skewed returns) and value (associated with distressed companies). Of course everything fell apart once the low-volatility anomaly was discovered, since there is no way to rationalize low risk companies as actually having more risk.
There are many reasons why publicly listed securities should be priced more efficiently than physical assets. They have low transaction costs, are perfectly fungible, free to transport, and all information is publically availaible. A comparison to real estate or employment isn't fair because of how sticky/illiquid those markets are, and because everyone has varying preferences.
The interesting thing about EMH is that it seems reasonable at first, then you realize it can't possibly reflect reality, but the harder you look the more unavoidable it becomes. The fact that long-term outperformance is so hard to achieve (especially on a risk-adjusted basis) shows that markets must be pretty damn efficient. Note that EMH only says that prices must reflect availaible information, not that they represent the true value of an asset at all points in time.