Restarting depletion calculation, drastic change in Keff

I am doing a depletion calculation on a reactor. Time step is one hour for 10 hour run, and I am seeing all the behaviors I expect in the first 10 hours - linear depletion of U235, Xe-135 build up, Pu-239 buildup, etc. Keff for each run of the integrator (one hour time steps) are:
array([[1.19590633, 0.00359115],
[1.18828691, 0.0028344 ],
[1.18583255, 0.00376955],
[1.17696794, 0.00301451],
[1.17378682, 0.0034014 ],
[1.1713465 , 0.00350791],
[1.16525232, 0.00368976],
[1.15468208, 0.00383944],
[1.16157056, 0.00319122],
[1.15115373, 0.00316935],
[1.15587757, 0.00309429]])

This all makes sense. I am trying to “extend” this run for another 10 hours using the following code:
mat=results.export_to_materials(10)
mat.cross_sections=“/home/brian/OpenMC_data/NuclearData/lib80x_hdf5/cross_sections.xml”
mat.export_to_xml()

model2 = openmc.Model(geometry=geometry,materials=mat, settings=settings)
operator2 = openmc.deplete.Operator(model2,“/home/brian/OpenMC_data/chain_casl_sfr.xml”)
integrator = openmc.deplete.PredictorIntegrator(operator2, time_steps, power)
integrator.integrate()
results2 = openmc.deplete.ResultsList.from_hdf5(“./depletion_results.h5”)
time, k = results2.get_eigenvalue()
time /= (60 * 60) # convert back to days from seconds
k

I think I am getting the material summary from hour 10 with the “mat=results.export_to_materials(10)” statement. But the initial Keff in the next burn calculation is wildly different than the final Keff of the first burn calcuation. The next iteration is Keff=0.8382

What am I doing incorrectly? Thanks!

============================> RESULTS <============================

k-effective (Collision) = 0.83797 +/- 0.00290
k-effective (Track-length) = 0.83826 +/- 0.00295
k-effective (Absorption) = 0.83933 +/- 0.00349
Combined k-effective = 0.83832 +/- 0.00233
Leakage Fraction = 0.01190 +/- 0.00048