Indeed TEMP(1) is current temperature (at the end of the increment). There's also TEMP(2) which is temperature increment. But if your subroutine doesn't work properly then there might be some error in the code. It's really hard to find exemplary Abaqus subroutines, especially for phase transformation, but there is one code attached to the paper titled "Two-dimensional simulation of thermal cutting of low-alloyed steels" by A. Laitinen from Tampere University of Applied Sciences.

Dear FEA way,
In the code, I tried with the thermal stress alone and it gives me correct output. Also, the code runs fine with austenite transformation. For checking the code I just created a simple model with 1500 predefined temperature and tried to cool down to room temperature. The expansion coefficient was given as subroutine (ie the equation for martensite expansion)In the subroutine if I change the temp(1) to a numerical value say for example 200/300 degree Celcius I will get exact analytical stress at this temperatures.

SUBROUTINE UEXPAN(EXPAN,DEXPANDT,TEMP,TIME,DTIME,PREDEF,
1 DPRED,STATEV,CMNAME,NSTATV,NOEL)
C
INCLUDE 'ABA_PARAM.INC'
C
CHARACTER*80 CMNAME
C
DIMENSION EXPAN(*),DEXPANDT(*),TEMP(2),TIME(2),PREDEF(*),
1 DPRED(*),STATEV(NSTATV)
REAL alpha,P,G,L,K,E,MS

MS=375

P=TEMP(1)-MS
K=0.011*P
E= EXP(K)
L=(-0.011)*E

G=(3.75)*(10.**(-3))*1.17*L*TEMP(2)


EXPAN(1)= G

RETURN
END