NaNs returned with SEP Aperture Photometry

[jortiz12]

Dear Community,

I'd appreciate help on the following issue.
Im doing apperture photometry with direct local background subtraction using the code

flux_lcbkg, fluxerr_lcbkg, flag_lcbkg = sep.sum_circle(datac_c, posx, posy, R, bkgann=(R_in, R_out), subpix=0)

, with R = 6, R_in = 32., R_out = 46. (pixels)

However, when using 1-channel images (from a MUSE datacube) I get many NaNs for the estimated flux (see bottom image below).
On the contrary, when using coadded 3-channel images (ie adding three wavelength bins along the spectral dimension of thre datacube), I don´t get any NaNs.

Any suggestion why I'm getting NaNs when using the former??? Plots of both images shown below.

Txs vm in advance
Jose