Mapping FIP ROI columns to fibers and brain regions in public dynamic-foraging data

Hi! We're reusing the public FIP dataset (aind-open-data, dynamic foraging and
Pavlovian sessions) for a methods study on photometry preprocessing, and we've hit
an ambiguity we can't resolve from the deposited files.

**Question:** For a given session, is there a mapping from the ROI/column index in
`FIP_DataG/Iso/R_*.csv` to (a) the connected patch cord / fiber, and (b) the brain
region and hemisphere that fiber targets?

**What we've already checked:**
- `session.json` declares the number of connected fibers but names them
`Fiber_0` / `Fiber_1` with no region attached.
- `FIP_ROIsG-Iso_*.csv` / `FIP_ROIsR_*.csv` contain CMOS ROI masks (roi index,
vertex index, x, y), not a fiber or region mapping.
- `procedures` gives injection targets, but not a per-fiber assignment.
- We're currently inferring which columns are real fibers by signal SD, which
resolves to inconsistent indices across sessions (0|1, 2|3, 0|2).

Is such a mapping deposited somewhere we've missed, or is the column ordering
guaranteed stable across sessions? We also noticed the processed pipeline NWB uses
`G_0`…`G_3` / `Iso_0`…`Iso_3` / `Red_0`…`Red_3` naming — does that fiber index
correspond directly to the raw CSV column index?

Two smaller related questions while we're here:
1. For Pavlovian sessions, is the expected reward-delivery time on omission trials
stored anywhere, or should it be derived from trial onset + median reward latency?
2. For dual-sensor animals, is there authoritative metadata assigning each sensor to
its excitation channel (green 470 / red 565), or is `procedures` the only source?

Happy to contribute documentation back if useful. Thanks for making this data public, the metadata is richer than most, which is why these gaps stand out.