Your LM317 also appears to be wired as a current source, and not as a voltage source. That aside...
THe DC-average of a rectified 120Vrms sine-wave is:
0.636 * 120Vrms = 76.632VDC.
And losses from clamping/linear regulation is
Voltage Wasted * Current = (76.632VDC - 50VDC) * 1.5A = 40W
THat would mean that between the resistor and LM317 you would need to dissipate 40W. Which basically means a 40W resistor since LM317's heat dissipation is almost neglible compared to 40W. You also need a zener that can handle 1.5A which is hard to find. Use a transformer between the wall and rectifier to greatly reduce this difference.
Also, having the smoothing capacitor after the voltage clamper causes the zener to constantly switch on and off (or rather, between blocking and breakdown). THat's why you have the smoothing capacitor before the voltage clamp. THe way you have it would make noise and other undesireable artifacts.
I would realy add in a transformer for safety sakes and efficiency . It's also one of the best things about AC conversion to increase efficiency so you might as well use it. Use it to to step down your AC voltage so that when you rectify the wave, the DC average is much closer to 50V which would improve efficiency of the clamp circuits and regulator.
0.636 * VACrms/n = VDC, where the transformer input : output turns ratio is n:1.
(The transformer acts on the AC voltage before rectification which is why division by n only applies to VACrms, and not 0.636*VACrms as a whole.)
A 1.4:1 or 1.5:1 transformer is about right to get close to 50VDC rectified (52VDC-55VDC to be exact) while still being high enough for the clamping circuits and linear regulator to deal with. It'd cut your power dissipation down to about 7-10W or so.
Transformer ratio of 1:2 would get you around 40V. Ratio of 1:3 would give you around 25V.
So what costs more? A 40W 100ohm resistor? Or a 100W transformer? The resistors are certainly rarer. Also consider that when when run at their power rating, resistors run hot enough to burn you and other things. One way around this is an resistor with an oversized power rating, but then you might as well spend the extra dough on a transformer and get better efficiency.
What is your load anyways? DOes it even need good regulation? Or even DC for that matter? If it didn't need regulation, that would make your life a lot easier because cooling can still be difficult at 10W.
EDIT: For example:
**broken link removed**
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=241-8-56-ND
I see what you mean by the cost for transformers is a lot more than resistors. But for me, non-isolation of the mains and the burning potential of the resistors is not acceptable. If they are acceptable to you, then you would have to first see if you can actually find such a resistor in low quantities.
And it seems that considering the cost of the transformer, you might as well just get one of these:
**broken link removed**