Getting SPD/ALT constraints correct in the FMC

OK, I am making progress with the FMC. However, I've now done two flights where I accepted the FMC generated SPD/ALT constraints for an ILS approach, i.e, where the data appeared in small fonts and was dynamic during the flight; and in both cases, at some point after the commencement of the descent, I observed that the two approach altitude constraints were wiped out and replaced with altitude constraints that were exactly the same as the penultimate waypoint alt prior to landing, i.e., about 1000 and 2000 vertical feet lower than they should have been.When I re-set the altitude constraint during such a descent today, the vertical descent indicator flipped out with newly calculated data once again and this time I received a message saying the descent was unachievable. I was rather amazed to see such a message since I was aiming now to descend a LESSSER vertical distance than a few moments earlier when the erroneous altitude constraint data had appeared on the CDU.Despite this and despite being suddenly and apparently almost 2000 vertical feet above the FMC calculated vertical descent path at around FL230, through constant monitoring and use of the speedbrakes, I was able somehow to stay in VNAV and meet the altitude and speed constraints and autoland completely successfully. VNAV did disconnect below 10,000 because my speed remained at the FMC descent speed of IAS 285, but I slowed to IAS 240 perfectly by around 6000' with the spoilers and reconnected to VNAV all the way down to the GS intercept.This was fantastic. I was impressed (I am learning....! )Anyway, my question is this:It seems to me that it may be necessary to set specific ("hard"?) SPD/ALT restrictions for the approach otherwise approach waypoints will have this FMC generated dynamic data and any such altitudes might be liable to just fall out of the flightplan as they have done twice. Is this assumption correct?I appreciate any advice or info on this item, thanks.Jonathan