That one looks good, apart from the tolerance (20%). It might be your best bet though.
FYI here's the schematic and some traces from a simulation to detect the average (mean) zener current.
In the schematic I have replaced the zener with a reverse-connected diode (which takes the "negative" current from the input capacitor), and a 27V zener in series with a diode and a shunt resistor. The voltage across the shunt resistor represents the current flowing through the zener when it is in its zener breakdown state, i.e. when the cathode is positive relative to the anode. This voltage is smoothed (averaged) by the 1K resistor and the 100 uF capacitor and measured on the "Vc" node.
I've temporarily removed the relay coil load by changing the coil resistance to 999,999 ohms, because this forces the zener to absorb more current.
The green trace (using the current scale on the right side of the graph) represents the current through the parallel diode, which peaks at about 160 mA. This current is actually a "negative" current but is shown as positive because LTSpice displays the conduction current as positive when you select a diode to monitor current flow.
The blue trace (again using the current scale on the right) represents the current through the zener. As you can see, it only passes current when the current from the input capacitor is positive. This occurs when Phase is rising relative to Neutral and current flow (conventional current) is forwards through the input capacitor. This current peaks at about 110 mA. (Again the trace is upside down.)
The red trace shows the voltage across Cc, the averaging capacitor, which is the average (mean) voltage across Rc, which represents the zener current. It levels out at around 26 mV, showing that the mean zener current is around 26 mA.
You multiply that current by the zener voltage, 27V in this case, to get the mean dissipation. 0.026 * 27 is 700 mW. With the input capacitor changed from 2.2 uF to 2.7 uF, this current rises to 36 mA which is nearly 1W dissipation. This is why I said that the input capacitor value is quite critical, which was true even though my calculations for the best value were wrong.
The reason I'm not interested in the current that flows when the zener cathode is negative is that during this time, the zener conducts in the forwards direction, like a diode, and its forward voltage is only ~0.7V so it dissipates very little power.