Hazard ratio for different groups with PHREG

[Lulu09]

There are 3 groups. And I want to get the hazard ratio for group 1 vs group 3, and group 2 vs group 3. The results that are generated from below 2 methods are not the same. So should I better put all 3 groups into PHREG, and set group 3 as reference level, Or should I select 2 groups each time when I execute PHREG procedure ? Thank you very much.

Method 1: 

proc phreg data = xx alpha=0.05;
    class trt (ref="3");
    model aval*cnsr(1)=trt /ties=exact;
    hazardratio trt / diff=all;
run;

Method 2: 

proc phreg data = xx(where=(trt in (1,3) )) alpha=0.05;
    class trt (ref="3");
    model aval*cnsr(1)=trt /ties=exact;
    hazardratio trt ;
run;

proc phreg data = xx(where=(trt in (2,3) )) alpha=0.05;
    class trt (ref="3");
    model aval*cnsr(1)=trt /ties=exact;
    hazardratio trt ;
run;

[OsoGris]

I think it is advisable to keep all three treatments in one model. If you want the differences with the reference category you can use the option DIFF=REF on the HAZARDRATIO statement.