Issues with Per-pixel plots

[orionlee]

I ran LATTE for the following TIC, and believe there are some issues with per-pixel plot output.

TICID,sectors,transits,BLS,model,FFI
128703021,"11,38","1601.85,2342.27",no,0,no

For sector 38 per-pixel plot,

• The indicated column / row is incorrect. The column is curiously similar to the one reported in sector 11. Based on TPF's header (1CRV4P, 2CRV4P), the (column, row) is (114, 101)
• the per-pixel plot seems to be very different from the actual per-pixel flux values. I don't know if it is a feature, the intended result of the LATTE-specific processing, or it is a bug.

Sector 38 per-pixel plot from LATTE:

Now compare it with the similar per-pixel plot produced by lightkurve:

• it looks very different from LATTE output.
• the per-pixel lightcurve is processed with a 5-sigma clipping of outliers, the default in lightkurve. I also plotted the flux as-is, without any processing, it looks similar.
• I also inspect some pixels independently. It appears that the plot from lightkurve reflects the actual flux.
  

I initially ran into the problem in LATTE-dev branch dated Oct 2020 (commit 5eea005), but is reproduced in the latest master.