How to tackle the r1400 error message with xtmixed

[Kamal Preet]

HELLO<

I have repeated measures data where patient scores are measured over four visits. The time gap between the four measurements is not equidistant, Hence, wanted to use the exponential covariance structure for the residuals with actual time in days rather than the visits. Briefly, the data description is as follows
Dependent variable - TotalUnmetStaff

Visit - counterSwH_OM10_13_NM4

Time - LOS_SwH11_14_days2 ( in days)

Patient - UniqueClientID_A3

The model with visits instead of the actual time gap between the subsequent visits converges

mixed TotalUnmetStaff ib(4).counterSwH_OM10_13_NM4 || UniqueClientID_A3:, noconstant var reml residuals(exponential, t(counterSwH_OM10_13_NM4))

however model with the actual time gap gives me error message

. mixed TotalUnmetStaff ib(4).counterSwH_OM10_13_NM4 || UniqueClientID_A3:, noconstant var reml residuals(exponential, t(LOS_SwH11_14_days2))

Obtaining starting values by EM:

Performing gradient-based optimization:

initial values not feasible
r(1400);

Please advice.