Simple calculations suggest this story is nonsense.
I'm not an expert in radiative physics but I can see from simple calculations that the story has strayed beyond the bounds of plausibility. The first, very simple, calculation is to look at the total power used if 200 mobile phones (a very full train carriage) were being used at the same time. I'm going to summarise from a couple of sources (the UK's Health Protection Agency has a good summary of the key issues and others are covered on www.antenna-theory.com). Point number 1 is that the worst case power transmission from a GSM phone is about 2W. So even here we can see that the total power emitted by 200 mobile phones is less than half that of a typical microwave (on the order of 1kW). Moreover, as @vartec points out in the comments, the train carriage has more than 10,000 times the volume of a microwave oven, so the specific effects on the contests will be a lot lower even if we assume all the energy is reflected back into the carriage (which is ridiculous not least because, if it were true, you wouldn't be able to use your phone in the carriage).
The second, more sophisticated, argument involves recognising that the peak power isn't the average power output. As the HPA site explains:
GSM mobile phones transmit their radio signals as 217 bursts of information every second. There is one burst every 4.6 ms (thousandth of a second) and each burst is 577 µs (millionths of a second) in duration. This means that, on average, they transmit for 1/8 of the time and their average output power is 8 times less than their peak output power.
Exposure guidelines, such as those published by ICNIRP, require exposures to be averaged over 6 minutes for comparison with their basic restrictions and it is more relevant to consider the average output power than the peak output power from phones. In this respect, GSM phones transmitting at 900 MHz and 1800 MHz have maximum time-averaged output powers of 0.25 W and 0.125 W respectively.
Another factor is also relevant here. GSM phones don't usually transmit at close to full power. Again the HPA summarise the reality well:
A key feature of mobile phone technology is that a mobile phone does not operate with a fixed output power level when a call is made. The maximum power output from a GSM mobile phone is around 2 W peak, but this can reduce in a sequence of 15 steps down to around 2 mW during calls, a power reduction factor of 1000.
So the typical power output is likely (crudely) 1,000 times lower than the reported peak (factor of 10 for time averaging factor of 100 for not always using maximum power) even ignoring the volume issue.
And then we have to consider where the power is absorbed. The antenna-theory site summarises part of the issue like this (not using quite the numbers or adjustments above):
The antenna is radiating, but less than half the power will be directed at your head - most radiates in all directions away. In addition, the antenna efficiency will be 50% for a good antenna that is held directly up against a head (the head actually detunes the antenna and makes it less efficient). Hence, of the 0.5 W of output power the phone transmits, there is a loss of at least 50% for the antenna efficiency, and at least 50% for the radiation that is not directed to your head. Hence, we can safely take 0.125 W (=0.5*0.5*0.5) as an upper bound for the power absorbed by your head.
This argument needs to be extended a little to cope with the 200-users-on-a-train scenario. But the key point it that the relevant thing for the amount of energy you will absorb is related to the distance from the transmitter and the cross sectional area of whatever is absorbing the radiation. If you are not holding the phone to your head, the possible absorption is much, much lower than the loss factor of 80% you get when you are.
So, again very crudely, if we start with a peak power output of 400W (200*2W) we should realistically factor in reductions of 10,000 (volume), 1,000 (average power) and perhaps 5 (cross section for absorption) to give a reduction of 50 million on the 400W initial estimate (which is half the typical Microwave's output). Or about 4 millionths of a typical microwave's intensity per unit volume. And the biggest contributor will be the phone held next to your head which isn't going to dump much more than 0.1W into you. If you are on a train outdoors where sunlight is shining through the window it might be worth putting this in context as antenna-theory does:
Is this a lot? Well, it is tough to say without comparison to something else. Let's take our good friend, sunlight on the Earth's surface. The power density of sunlight is roughly 1.35 kW/m^2 (killiWatts [sic] per square meter). Now, the bigger your head is, the more energy your head absorbs from the sunlight. Let's say you have a standard adult head, which we'll approximate with a circular cross section of radius 4". Your head would then be roughly 0.0324 square meters in cross section. As a result, the power absorbed by your head will be roughly 1.35*0.0324 = 0.0438 kW = 43.8 W.
You might also want to consider the fact that that the 40-odd watts of sunlight contains ionising UV which is a known carcinogen.