I don’t think it’s that simple. Disclaimer: I am not an expert in lossy data compression
Lossy video CODECS don’t just compress each frame a la jpeg in sequence and string them together, they look at changes from frame to frame and use patterns there too, to compress its data (that’s why with really heavy video compression, you can see artefacts at scene change boundaries). This is needed to further reduce the size/bit rate enough for consumer transmission/use/storage.
I don’t think you can win here without, as you say, using lossless compression or uncompressed formats for your editing stages. The only (no further) lossless manipulation I know of is lossless jpeg rotation where the pattern algorithms must be applicable whichever way round the image is, or can be translated or manipulated in “pattern space” to result in no further decompression/compression losses.
Video compression is across frames, so even if you edit at frame boundaries and treat frames as distinct independent images, your video CODEC doesn’t. Sorry to be the bearer of bad news.
>