The heart rate monitor seems to work pretty well while cycling. You can correlate the increase in rate to individual hills, so I doubt that it's doing much guessing/estimating.
Ah ... but you're guessing that its not guessing/estimating ;-)
These things use pulse oximetry or photoplethysmography techniques. If you've ever looked at the signal that you get from the receiving photodetector and seen all the noise/interference that you get, you'd be amazed that these devices work at all!
Significant filtering and digital signal processing has to be done to to extract the heartbeat from all this noise. They can do this when the SN ratio is reasonable (i.e. you're fairly stationary), but movement with all sorts of bumps and jars etc causes a significant amount of additional noise, lots of it semi repetitive (like a heartbeat!) so it becomes much, much more difficult for the devices to work out what's the relevant signal in all that jumble. Hence a lot of the time, they have to extrapolate, guess based on previous readings etc - depends on the algorithms.
But it is well known that these wrist worn devices are not particularly accurate when doing exercise (and I stress again - while doing exercise - they are fine for resting heartbeat)
In fact, I'm fairly sure there was (is?) a class action lawsuit going on in the US against fitbit because of this inaccuracy.
Bottom line - if you really want to know what your pulse is doing when exercising, you can't trust a wrist based sensor, and you have to rely on the chest strap type.
Which is unfortunate, because the wrist devices are so convenient, and many people (myself included!) really dislike wearing the chest straps!